Cable soft fault diagnosis method and device, computer equipment and readable storage medium

Abstract

The application provides a cable soft fault diagnosis method and device, computer equipment and a readable storage medium, and relates to the technical field of power equipment. The method comprises the following steps: collecting first TDR response data of a fault-free cable at multiple detection frequencies, and constructing a PCA reference model based on the first TDR response data at the multiple detection frequencies; collecting second TDR response data of a to-be-tested cable at the same multiple detection frequencies, and determining whether the to-be-tested cable has a soft fault based on the second TDR response data and the constructed PCA reference model, wherein the PCA reference model is used to determine whether the to-be-tested cable has a soft fault by calculating a squared prediction error statistic; and if it is determined that the to-be-tested cable has a soft fault, the contribution of each detection frequency to the squared prediction error statistic is analyzed to determine a target detection frequency.

Description

Technical Field

[0001] This application relates to the field of power equipment technology, and in particular to a method, apparatus, computer equipment, and readable storage medium for diagnosing cable soft faults. Background Technology

[0002] With the growth of electricity demand and the integration of distributed energy, cables are subjected to multiple stresses during operation, including vibration, high temperature, electromagnetic interference and mechanical wear. This causes minute deformation and material deterioration in the cables, leading to soft faults such as shielding layer damage and insulation layer micro-cracks. These defects may initially have weak signals, but they can gradually worsen and eventually cause system paralysis, resulting in significant losses.

[0003] The cable fault diagnosis methods in related technologies have the following problems: (1) Unreasonable selection of detection frequency: When the detection frequency is manually selected, the high-frequency signal attenuates quickly in the cable and the low-frequency signal has insufficient resolution, making it difficult to balance the detection depth and accuracy; and the use of single-frequency detection does not consider the influence of the fault location on the signal response, nor does it deeply analyze the variation law and correlation of reflected signals under multiple frequencies, making it unable to adapt to different fault scenarios. (2) Low reliability of fault judgment: The fault judgment method based on experience thresholds does not fully explore the characteristic correlation of reflected signals, and is prone to misjudgment and missed judgment in complex environments, resulting in insufficient reliability of the diagnostic results and difficulty in meeting the needs of safe cable operation. (3) Weak ability to identify soft faults: Traditional reflection methods are difficult to identify small impedance changes caused by soft faults such as shielding layer damage and insulation layer micro-cracks. The initial signal of soft faults is weak and cannot effectively detect such faults.

[0004] Therefore, how to achieve accurate diagnosis of cable soft faults has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, this application provides a method, apparatus, computer equipment, and readable storage medium for diagnosing cable soft faults, enabling accurate diagnosis of cable soft faults.

[0006] In a first aspect, embodiments of this application provide a method for diagnosing cable soft faults, including:

[0007] Collect first TDR (Time Domain Reflectometry) response data of fault-free cables at multiple detection frequencies, and construct a PCA (Principal Component Analysis) benchmark model based on the first TDR response data at multiple detection frequencies;

[0008] Second TDR response data of the cable under test are collected at multiple detection frequencies. Based on the second TDR response data and the established PCA baseline model, it is determined whether there is a soft fault in the cable under test. The PCA baseline model is used to determine whether there is a soft fault in the cable under test by calculating the squared prediction error statistic.

[0009] If a soft fault is determined in the cable under test, the contribution of each detection frequency to the squared prediction error statistic is analyzed to determine the target detection frequency.

[0010] The method described in the embodiments of this application may also have the following additional technical features:

[0011] In the above technical solution, optionally, the step of collecting first TDR response data of fault-free cables at multiple detection frequencies and constructing a PCA benchmark model based on the first TDR response data at multiple detection frequencies includes:

[0012] For fault-free cables, Gaussian pulse signals are injected using TDR technology at multiple preset detection frequencies, and the first TDR response data corresponding to each detection frequency is recorded.

[0013] A PCA baseline model is constructed based on the first TDR response data at multiple detection frequencies. The PCA baseline model includes a data processing unit, a principal component extraction unit, and a fault judgment unit. The data processing unit is used to preprocess the first TDR response data. The principal component extraction unit is used to extract principal components based on the preprocessed data. The fault judgment unit is used to calculate the squared prediction error statistic based on the principal components, set the fault judgment threshold, and perform soft fault judgment based on the squared prediction error statistic and the fault judgment threshold.

[0014] In the above technical solution, optionally, the data processing unit is specifically used to: construct a first health matrix based on the first TDR response data corresponding to each detection frequency, and normalize the elements in the first health matrix so that the first TDR response data at each detection frequency are at the same order of magnitude, thereby obtaining a second health matrix;

[0015] The principal component extraction unit is specifically used to: perform singular value decomposition on the second health matrix to obtain a left singular vector matrix, a singular value diagonal matrix, and a right singular vector matrix. The right singular vector matrix is ​​used to determine each principal component, which is a linear combination of the first TDR response data. The left singular vector matrix is ​​used to determine the numerical score of the principal component. The singular value diagonal matrix is ​​used to determine the variance contribution of the principal component. The unit also sorts each principal component according to its variance contribution from high to low and retains the top k principal components, where k is a positive integer.

[0016] The fault judgment unit is specifically used to: calculate the squared prediction error statistic based on the retained k principal components, set a fault judgment threshold, and determine that the cable has a soft fault when the squared prediction error statistic is greater than or equal to the fault judgment threshold.

[0017] Optionally, in the above technical solution, the step of collecting second TDR response data of the cable under test at multiple detection frequencies, and determining whether the cable under test has a soft fault based on the second TDR response data and the established PCA benchmark model, includes:

[0018] For the cable under test, Gaussian pulse signals were injected using TDR technology at multiple detection frequencies, the same as those used in the model training phase, and the second TDR response data corresponding to each detection frequency was recorded.

[0019] Based on the second TDR response data corresponding to each detection frequency, a first detection matrix is ​​constructed;

[0020] The parameters of the normalization process during the model training phase are used to normalize the elements in the first detection matrix so that the second TDR response data at each detection frequency are on the same order of magnitude, thus obtaining the second detection matrix.

[0021] The second detection matrix is ​​input into the constructed PCA baseline model to obtain the result of whether the cable under test has a soft fault.

[0022] In the above technical solution, optionally, each column of the first detection matrix and / or the second detection matrix is ​​a column vector under a detection frequency, and the elements in the column vector include TDR response data of multiple sampling points corresponding to multiple fault scenarios characterized by fault location, fault length and fault width.

[0023] Optionally, in the above technical solution, inputting the second detection matrix into the constructed PCA benchmark model to obtain the result of whether the cable under test has a soft fault includes:

[0024] The second detection matrix is ​​input into the constructed PCA baseline model to obtain the squared prediction error statistics for different sampling points;

[0025] If the squared prediction error statistic corresponding to multiple consecutive sampling points is greater than or equal to the fault judgment threshold, then the cable under test is determined to have a soft fault.

[0026] In the above technical solution, optionally, the step of analyzing the contribution of each detection frequency to the squared prediction error statistic to determine the target detection frequency if a soft fault is determined in the cable under test includes:

[0027] For each sampling point in the fault area of ​​the cable under test where there is a soft fault, a contribution analysis is performed to calculate the contribution of each detection frequency to the squared prediction error statistic.

[0028] The detection frequency with the highest contribution is selected as the target detection frequency.

[0029] Secondly, embodiments of this application provide a cable soft fault diagnosis device, comprising:

[0030] The model building module is used to collect the first TDR response data of the fault-free cable at multiple detection frequencies, and to build a PCA benchmark model based on the first TDR response data at multiple detection frequencies.

[0031] The soft fault diagnosis module is used to collect the second TDR response data of the cable under test at multiple detection frequencies. Based on the second TDR response data and the established PCA benchmark model, it determines whether there is a soft fault in the cable under test. The PCA benchmark model is used to determine whether there is a soft fault in the cable under test by calculating the squared prediction error statistic.

[0032] The frequency determination module is used to analyze the contribution of each detection frequency to the squared prediction error statistic if a soft fault is determined in the cable under test, in order to determine the target detection frequency.

[0033] Thirdly, embodiments of this application provide a computer device including a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions implementing the steps of the method as described in the first aspect when executed by the processor.

[0034] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method as described in the first aspect.

[0035] This application presents a cable soft fault diagnosis method, apparatus, computer equipment, and readable storage medium, proposing a cable soft fault diagnosis scheme based on joint TDR and PCA detection. The scheme achieves cable soft fault diagnosis through a three-stage process: training, monitoring, and frequency selection. TDR response data of fault-free cables at multiple frequencies are collected and normalized to construct a PCA baseline model. TDR data of the cable under test at the same frequency are collected, normalized, and projected onto the PCA baseline model. The Q-value is used to determine whether a soft fault exists in the cable under test. If a soft fault is detected, the contribution of each frequency to the statistics is analyzed, and a target detection frequency is selected, ultimately achieving accurate soft fault diagnosis and optimization of detection parameters.

[0036] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0037] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0038] Figure 1 One of the flowcharts of a cable soft fault diagnosis method according to an embodiment of this application is shown;

[0039] Figure 2 A second schematic flowchart of the cable soft fault diagnosis method according to an embodiment of this application is shown;

[0040] Figure 3 A schematic diagram showing the Q value of the experimentally measured sample according to an embodiment of this application is illustrated;

[0041] Figure 4 A schematic diagram illustrating the frequency contribution of the F9 fault according to an embodiment of this application is shown;

[0042] Figure 5 A schematic diagram of the time-domain reflection (TDR) response of a shielding layer failure according to an embodiment of this application is shown;

[0043] Figure 6 A structural block diagram of a cable soft fault diagnosis device according to an embodiment of this application is shown;

[0044] Figure 7 A structural block diagram of a computer device according to an embodiment of this application is shown. Detailed Implementation

[0045] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0046] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0047] The cable soft fault diagnosis method, apparatus, computer equipment, and readable storage medium provided in this application will be described in detail below with reference to the accompanying drawings and through specific embodiments and application scenarios. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0048] This application provides a method for diagnosing soft faults in cables. Figure 1 and Figure 2 A flowchart illustrating a cable soft fault diagnosis method according to an embodiment of this application is shown. The method includes:

[0049] Step 101: Collect the first TDR response data of the fault-free cable at multiple detection frequencies, and construct the PCA benchmark model based on the first TDR response data at multiple detection frequencies.

[0050] In this step, based on the TDR response data of the fault-free cable, i.e., the healthy cable, at multiple detection frequencies, a PCA benchmark model of the fault-free cable is constructed to provide a comparison benchmark for subsequent fault detection.

[0051] In one embodiment of this application, the acquisition of first TDR response data of a fault-free cable at multiple detection frequencies, and the construction of a PCA benchmark model based on the first TDR response data at multiple detection frequencies, includes:

[0052] For fault-free cables, Gaussian pulse signals are injected using TDR technology at multiple preset detection frequencies, and the first TDR response data corresponding to each detection frequency is recorded.

[0053] A PCA baseline model is constructed based on the first TDR response data at multiple detection frequencies. The PCA baseline model includes a data processing unit, a principal component extraction unit, and a fault judgment unit. The data processing unit is used to preprocess the first TDR response data. The principal component extraction unit is used to extract principal components based on the preprocessed data. The fault judgment unit is used to calculate the squared prediction error statistic based on the principal components, set the fault judgment threshold, and perform soft fault judgment based on the squared prediction error statistic and the fault judgment threshold.

[0054] In one embodiment of this application, the data processing unit is specifically used to: construct a first health matrix based on the first TDR response data corresponding to each detection frequency, and normalize the elements in the first health matrix so that the first TDR response data at each detection frequency are at the same order of magnitude, thereby obtaining a second health matrix;

[0055] The principal component extraction unit is specifically used to: perform singular value decomposition on the second health matrix to obtain a left singular vector matrix, a singular value diagonal matrix, and a right singular vector matrix. The right singular vector matrix is ​​used to determine each principal component, which is a linear combination of the first TDR response data. The left singular vector matrix is ​​used to determine the numerical score of the principal component. The singular value diagonal matrix is ​​used to determine the variance contribution of the principal component. The unit also sorts each principal component according to its variance contribution from high to low and retains the top k principal components, where k is a positive integer.

[0056] The fault judgment unit is specifically used to: calculate the squared prediction error statistic based on the retained k principal components, set a fault judgment threshold, and determine that the cable has a soft fault when the squared prediction error statistic is greater than or equal to the fault judgment threshold.

[0057] In this embodiment, a fault-free cable of the same type as the cable under test is selected. Gaussian pulse signals are injected using TDR technology at multiple preset detection frequencies, such as 1MHz, 2MHz, 3MHz, and 4MHz. The TDR response data corresponding to each frequency is recorded, i.e., discretized "distance-amplitude" sampling points. These data are then organized into a first health matrix. The data in the first health matrix are column vectors of health TDR response at various frequencies. The dimension is m×n, which is the number of TDR response sampling points × the number of frequencies.

[0058] Taking a frequency of 4 as an example, the first health matrix for:

[0059]

[0060] Where j = 1, 2, 3, 4, The column vector of the health TDR response at the j-th frequency.

[0061] To eliminate the interference of differences in TDR response amplitude range at different frequencies on the model, the first health matrix was modified. All data are normalized to ensure that the TDR response data at each frequency are of the same order of magnitude. The normalization formula is as follows:

[0062]

[0063] Where x is the first health matrix The data in the middle, mean and standard deviation It is the first health matrix The mean and standard deviation of the TDR response data at the corresponding frequency. This is the normalized data.

[0064] After normalization, we obtain the second health matrix X, which has dimensions m×n. We then perform singular value decomposition (SVD) on the second health matrix X:

[0065]

[0066] Where U is the left singular vector, an m×m orthogonal matrix, corresponding to the numerical score or score vector in principal component analysis, reflecting the projection of the data into the principal component space. It is an m×n diagonal matrix, with diagonal elements These are the singular values ​​of X, directly related to the variance contribution of the principal components. The variance contribution can be determined by... Calculation. V is the right singular vector, an n×n orthogonal matrix, corresponding to the loading vector in principal component analysis, reflecting the linear relationship between the original detection data (second TDR response data at each frequency) and the principal components.

[0067] It should be noted that principal component analysis is essentially a comprehensive extraction of the original test data. Its purpose is to remove redundant information and retain the core patterns to facilitate subsequent fault comparison. It does not directly present the original signal parameters, but rather extracts the common patterns of the original parameters to reflect the health status of the cable.

[0068] First, the raw test data consists of the TDR response signals of a healthy cable at multiple frequencies. For a healthy cable, a set of TDR response signals will be measured at each frequency, which can be understood as a signal curve containing information such as signal amplitude and propagation time. These signals are not completely independent. For example, the TDR response signals of high frequency (4MHz) and medium frequency (2MHz) both contain the impedance characteristics of the healthy cable and the propagation characteristics of the signal in the cable. This repetitive information is data redundancy.

[0069] Secondly, principal components (PCs) are a few uncorrelated composite variables obtained by mathematically reorganizing multiple sets of related raw data using the PCA algorithm. Simply put, it is not the original TDR response signal at a single frequency, but rather a weighted combination of multiple frequency raw signals according to their importance, extracting the features that best represent the essential characteristics of healthy cables. For example, performing PCA on four sets of frequency data for healthy cables yields four principal components: PC1, PC2, PC3, and PC4. However, their importance is completely different. The variance contribution represents how much variation in the original data this principal component can explain; the higher the variance contribution, the more core information it contains from the original data, and the more important it is.

[0070] Furthermore, by analyzing the variance contribution of each principal component, the principal components that can characterize the core features of the original detection data are retained, thereby achieving dimensionality reduction of high-dimensional data and preserving the TDR response pattern of healthy cables to the greatest extent.

[0071] Fault identification is performed using the Q-value (squared prediction error, SPE) in the residual subspace. The Q-value reflects the degree of deviation between the data and the principal component space, that is, the degree of deviation between the tested data and the PCA baseline model. It is a core indicator for identifying soft faults; the greater the deviation, the higher the probability of a fault. The formula for calculating the Q-value is:

[0072]

[0073] Where n is the number of columns in the matrix, and k is the number of principal components to be retained, which can be 1. The projection coefficients of the data in the residual subspace are the differences between the original data and the projections into the principal component space.

[0074] Set a fault judgment threshold as the upper control limit of the Q value, which can be calculated by multiplying the Q value by... , The threshold can be 95%; if it exceeds this fault judgment threshold, the data is considered abnormal.

[0075] Understandably, the PCA benchmark model is essentially a health status standard framework trained using TDR data from healthy cables. It consists of two core parts: (1) Principal component space: extracting the core patterns of health data, such as impedance stability and attenuation patterns; (2) Residual subspace: accommodating minor, insignificant differences in the health data. The role of the PCA benchmark model is to place the TDR data of the cable under test into this framework during subsequent testing. If the data conforms to the health patterns of the principal component space and the deviation is within the threshold range, it is considered healthy; if the deviation exceeds the threshold, there may be a fault.

[0076] The construction process of the PCA benchmark model includes a complete process of "data preparation → model construction → threshold determination". The core components and construction process of the PCA benchmark model are: (1) Data preparation, collecting and standardizing health data, which is the raw material processing of the model; (2) Model construction, extracting principal components using singular value decomposition, which is the core skeleton of the model; (3) Threshold determination, setting confidence levels and calculating the upper control limit of the Q value, which is the judgment criterion of the model.

[0077] Step 102: Collect the second TDR response data of the cable under test at multiple detection frequencies. Based on the second TDR response data and the established PCA benchmark model, determine whether there is a soft fault in the cable under test. The PCA benchmark model is used to determine whether there is a soft fault in the cable under test by calculating the squared prediction error statistic.

[0078] In this step, fault detection is performed on the cable under test. By comparing the TDR response data of the cable under test with the PCA baseline model, it is determined whether there is a soft fault in the cable under test.

[0079] In one embodiment of this application, the step of collecting second TDR response data of the cable under test at multiple detection frequencies, and determining whether the cable under test has a soft fault based on the second TDR response data and the established PCA benchmark model, includes:

[0080] For the cable under test, Gaussian pulse signals were injected using TDR technology at multiple detection frequencies, the same as those used in the model training phase, and the second TDR response data corresponding to each detection frequency was recorded.

[0081] Based on the second TDR response data corresponding to each detection frequency, a first detection matrix is ​​constructed;

[0082] The parameters of the normalization process during the model training phase are used to normalize the elements in the first detection matrix so that the second TDR response data at each detection frequency are on the same order of magnitude, thus obtaining the second detection matrix.

[0083] The second detection matrix is ​​input into the constructed PCA baseline model to obtain the result of whether the cable under test has a soft fault.

[0084] Each column of the first detection matrix and / or the second detection matrix is ​​a column vector at a detection frequency, and the elements in the column vector include TDR response data of multiple sampling points corresponding to various fault scenarios characterized by fault location, fault length and fault width.

[0085] In this embodiment, for the cable under test that may have soft faults such as shielding layer damage or compression, second TDR response data is collected at multiple detection frequencies, such as 1MHz, 2MHz, 3MHz, and 4MHz, the same as in the training phase, and organized into a first detection matrix. Its dimensions are similar to the first health matrix. Consistent. Each column within it. For example, a column vector of the TDR response to be measured at a detection frequency. , TDR response feature data corresponding to 36 possible fault scenarios characterized by three target parameters, including fault location. Fault length and fault width Fault location [20%, 50%, 70%, 90%], changing the fault location as a percentage along the cable length; fault length [5mm, 10mm, 20mm], physical length of the damaged area of ​​the shielding layer along the cable axis; fault width. [45°, 90°, 180°], the circumferential angle of shielding damage on the cable cross-section.

[0086] To ensure consistency with health TDR response data processing standards and avoid introducing additional errors, the mean calculated during the training phase is used. and standard deviation For the detection matrix Normalization is performed to obtain the second detection matrix. The data from the normalized second detection matrix is ​​then projected onto the PCA baseline model built during the training phase to obtain the result of whether the cable under test has a soft fault.

[0087] In one embodiment of this application, the step of inputting the second detection matrix into the constructed PCA baseline model to obtain the result of whether the cable under test has a soft fault includes:

[0088] The second detection matrix is ​​input into the constructed PCA baseline model to obtain the squared prediction error statistic (i.e., Q value) corresponding to different sampling points.

[0089] If the squared prediction error statistic corresponding to multiple consecutive sampling points is greater than or equal to the fault judgment threshold, then the cable under test is determined to have a soft fault.

[0090] In this embodiment, the second detection matrix is ​​input into the constructed PCA baseline model. The PCA baseline model calculates the Q value for each sampling point (i.e., each test sample), where one test sample is a row of multi-frequency data. The Q value is obtained by calculating the projection error of the data in the residual space. That is, the residual vector is obtained by subtracting the "healthy row data reconstructed by the PCA baseline model" from the "original row data", and then the sum of squares of the residuals is calculated to obtain the Q value.

[0091] The calculated Q value is compared with the fault judgment threshold set in the PCA baseline model. If the Q value does not exceed the fault judgment threshold, the cable is determined to be fault-free.

[0092] If multiple consecutive sampling points (i.e., multiple consecutive rows in the detection matrix, corresponding to consecutive spatial positions on the cable) all meet the condition that the Q value exceeds the fault judgment threshold, then the cable under test is determined to have a soft fault. This eliminates the possibility of a single sampling point exceeding the Q value due to accidental errors such as minor instrument fluctuations or brief environmental interference, ensuring the reliability of fault judgment.

[0093] Step 103: If it is determined that there is a soft fault in the cable under test, analyze the contribution of each detection frequency to the squared prediction error statistic in order to determine the target detection frequency.

[0094] In this step, for the detected soft faults, the target detection frequency that best highlights the fault characteristics is selected, which is the optimal detection frequency.

[0095] In one embodiment of this application, the step of analyzing the contribution of each detection frequency to the squared prediction error statistic to determine the target detection frequency if a soft fault is determined in the cable under test includes:

[0096] For each sampling point in the fault area of ​​the cable under test where there is a soft fault, a contribution analysis is performed to calculate the contribution of each detection frequency to the squared prediction error statistic.

[0097] The detection frequency with the highest contribution is selected as the target detection frequency.

[0098] In this embodiment, for sampling points identified as soft faults, contribution analysis is used to calculate the contribution of each detection frequency to the Q value. The higher the contribution, the better the TDR response at that detection frequency reflects the anomaly caused by the fault, such as more obvious fluctuations in the reflection amplitude at the fault location. The detection frequency with the highest contribution is selected as the target detection frequency.

[0099] Understandably, the contribution of each frequency to the Q-value is essentially achieved by decomposing the Q-value into each frequency dimension. The contribution of that frequency to the fault anomaly is measured by calculating the proportion of the squared residual component corresponding to a specific frequency to the total Q-value. A higher proportion indicates that the TDR response at that frequency more clearly highlights the fault characteristics. Calculation steps:

[0100] Step 1: For the fault region corresponding to consecutive out-of-range sampling points, merge the residual information of all out-of-range sampling points in the region to obtain the total Q value of the region.

[0101] For example: In a certain fault area, there are 3 consecutive sampling points exceeding the standard. The residual vector of each sampling point is:

[0102] Sampling point 1: [+0.02, +0.05, -0.08, +0.15]

[0103] Sampling point 2: [-0.01, +0.06, +0.09, +0.16]

[0104] Sampling point 3: [+0.03, -0.04, +0.07, +0.14]

[0105] For each frequency, sum the squared residuals of all sampling points within the fault region:

[0106] 1MHz: (0.02)²+(-0.01)²+(0.03)²=0.0014

[0107] 2MHz: (0.05)²+(0.06)²+(-0.04)²=0.0077

[0108] 3MHz: (-0.08)²+(0.09)²+(0.07)²=0.0194

[0109] 4MHz: (0.15)²+(0.16)²+(0.14)²=0.0677

[0110] Calculate the total Q value for the region:

[0111] 0.0014 + 0.0077 + 0.0194 + 0.0677 = 0.0962

[0112] Step 2: Calculate the contribution of a single frequency to the Q value

[0113] The contribution of a certain frequency is defined as the percentage of the square of the residual component at that frequency to the total Q value. The formula for calculating the contribution is:

[0114]

[0115] Among them, C j The contribution of the j-th frequency, It is the square of the j-th frequency residual component.

[0116] For example: the contribution of each frequency to the Q value:

[0117] 1MHz: (0.0014 / 0.0962)×100%≈1.46%

[0118] 2MHz: (0.0077 / 0.0962)×100%≈8.00%

[0119] 3MHz: (0.0194 / 0.0962)×100%≈20.17%

[0120] 4MHz: (0.0677 / 0.0962)×100%≈70.37%

[0121] It should be noted that e i It is the component of the residual in the "direction of the minor principal components", e j It represents the component of the residual in a "single frequency direction". The essence of the Q value is the "sum of squares of the residual vectors" (the residual vectors are fixed), therefore, regardless of the principal component dimension e... i The division is still based on the frequency dimension e. j The final total Q value obtained from the two decompositions is exactly the same; the two decompositions simply break down the same Q value from different perspectives, corresponding to different analytical objectives: Principal Component Decomposition (e... i ) serves "fault determination", frequency dimension decomposition (e j It serves as a tool for "optimal frequency screening".

[0122] The formula for calculating the total Q value can be:

[0123]

[0124] Step 3: Filter the optimal frequency

[0125] The frequency with the highest contribution is the target detection frequency, which best highlights the fault characteristics. This frequency can be used first to accurately detect or locate similar faults in the future.

[0126] For example, the optimal frequency selection: 4MHz has the highest contribution (70.37%), so 4MHz is selected as the target detection frequency for this fault.

[0127] This application demonstrates significant technical advantages, comprehensively overcoming the shortcomings of traditional methods, and achieving breakthroughs in "fault location". Fault length Angled tangent Taking the F9 soft fault case as an example, regarding the problem of "weak soft fault identification capability," traditional time-domain reflectometry is difficult to identify due to the weak soft fault signal. However, the method in this application fuses multi-frequency TDR response features through a PCA model. Figure 3 The Q value of the F9 fault sample significantly exceeded the health data control limit (threshold 0.628), accurately capturing the minute impedance changes caused by the soft fault. Addressing the issue of "inappropriate detection frequency selection," traditional manual frequency selection cannot balance depth and accuracy and has poor adaptability. The method in this application dynamically selects frequencies by analyzing the contribution of each frequency to the fault statistics. Figure 4 This indicates that the 4MHz frequency contributes the most, combined with Figure 5 As can be seen, the reflection peak characteristics at the fault location are clear at 4MHz, ensuring high resolution at high frequencies while balancing signal attenuation due to the proximity of the fault to the injection end, thus achieving a balance between detection depth and accuracy. Regarding the issue of "low reliability in fault diagnosis," traditional empirical thresholds are prone to misjudgment and missed detection. The method in this application uses Q-value determination to fully exploit the correlation of multi-frequency reflection signal characteristics, ensuring high reliability in fault diagnosis under complex scenarios. In summary, the method in this application, through PCA feature extraction and statistical analysis of multi-frequency TDR data, achieves accurate identification of weak fault signals, optimal frequency selection for fault location, and highly reliable fault diagnosis in the F9 soft fault scenario, comprehensively solving the technical problems of traditional methods.

[0128] This application proposes a three-stage fusion process of "training-monitoring-frequency selection", which first constructs a health matrix using multi-frequency TDR data. After normalization and singular value decomposition, high-variance principal components are retained to form a healthy PCA baseline model. Then, TDR data of the same frequency are collected from the cable under test to construct a detection matrix. This paper proposes a method for accurate diagnosis of cable soft faults by using the Q-value of the residual subspace to determine faults and fully explore the characteristic correlations of multi-frequency reflected signals, thus solving the problems of misjudgment and missed detection by traditional empirical thresholds. If a fault is detected, the contribution of each frequency to the statistic is analyzed, and the frequency with the highest contribution is selected as the target detection frequency, achieving a dynamic balance between detection accuracy and depth under different fault scenarios. This application constructs a scheme for accurate diagnosis of cable soft faults from three dimensions: "matrix processing of multi-frequency TDR data and PCA dimensionality reduction modeling", "fault determination mechanism of squared prediction error (Q-value) statistic", and "contribution-oriented dynamic optimal frequency selection method", breaking through the limitations of traditional methods in frequency selection and fault identification reliability.

[0129] It is worth noting that soft faults do not immediately interrupt cable signal or energy transmission like hard faults such as open circuits or short circuits, and are difficult to detect in the early stages. However, they gradually worsen with use (aging) and environmental influences (vibration, humidity, high temperature), eventually evolving into hard faults and causing the entire system to fail. This application precisely solves the pain points of weak signals, easy masking, and difficulty in frequency selection caused by soft faults through a combination of "reflection method + PCA + Q value + automatic frequency selection". Specifically:

[0130] (1) Use PCA to amplify the small differences of soft faults to solve the problem of weak signals being difficult to identify. PCA can compress complex signals at multiple frequencies into principal components, that is, core features. This is equivalent to amplifying the vague small differences into clear feature deviations. For example, the signal of a healthy cable will be concentrated in a normal area in the PCA model. Even if the soft fault causes only a small change in the signal, it will deviate from this area and be easily identified.

[0131] (2) Use Q value to sensitively capture anomalies and solve the problem of missed soft fault detection. The PCA model will predict what signal a healthy cable should have. If the error (Q value) between the actual signal and the predicted signal of the cable under test exceeds the threshold, it is determined that there is a fault.

[0132] (3) Automatically select the best frequency to solve the problem of difficulty in selecting frequency.

[0133] As a specific implementation of the above-mentioned cable soft fault diagnosis method, this application provides a cable soft fault diagnosis device. For example... Figure 6 As shown, the cable soft fault diagnosis device 600 includes: a model building module 601, a soft fault diagnosis module 602, and a frequency determination module 603.

[0134] Among them, the model building module 601 is used to collect the first TDR response data of the faultless cable at multiple detection frequencies, and build a PCA benchmark model based on the first TDR response data at multiple detection frequencies.

[0135] The soft fault diagnosis module 602 is used to collect the second TDR response data of the cable under test at multiple detection frequencies. Based on the second TDR response data and the established PCA benchmark model, it determines whether there is a soft fault in the cable under test. The PCA benchmark model is used to determine whether there is a soft fault in the cable under test by calculating the squared prediction error statistic.

[0136] The frequency determination module 603 is used to analyze the contribution of each detection frequency to the squared prediction error statistic if a soft fault is determined in the cable under test, so as to determine the target detection frequency.

[0137] Furthermore, the model building module 601 is specifically used for:

[0138] For fault-free cables, Gaussian pulse signals are injected using TDR technology at multiple preset detection frequencies, and the first TDR response data corresponding to each detection frequency is recorded.

[0139] A PCA baseline model is constructed based on the first TDR response data at multiple detection frequencies. The PCA baseline model includes a data processing unit, a principal component extraction unit, and a fault judgment unit. The data processing unit is used to preprocess the first TDR response data. The principal component extraction unit is used to extract principal components based on the preprocessed data. The fault judgment unit is used to calculate the squared prediction error statistic based on the principal components, set the fault judgment threshold, and perform soft fault judgment based on the squared prediction error statistic and the fault judgment threshold.

[0140] Furthermore, the data processing unit is specifically used to: construct a first health matrix based on the first TDR response data corresponding to each detection frequency, and normalize the elements in the first health matrix so that the first TDR response data at each detection frequency are of the same magnitude, thereby obtaining a second health matrix;

[0141] The principal component extraction unit is specifically used to: perform singular value decomposition on the second health matrix to obtain a left singular vector matrix, a singular value diagonal matrix, and a right singular vector matrix. The right singular vector matrix is ​​used to determine each principal component, which is a linear combination of the first TDR response data. The left singular vector matrix is ​​used to determine the numerical score of the principal component. The singular value diagonal matrix is ​​used to determine the variance contribution of the principal component. The unit also sorts each principal component according to its variance contribution from high to low and retains the top k principal components, where k is a positive integer.

[0142] The fault judgment unit is specifically used to: calculate the squared prediction error statistic based on the retained k principal components, set a fault judgment threshold, and determine that the cable has a soft fault when the squared prediction error statistic is greater than or equal to the fault judgment threshold.

[0143] Furthermore, the soft fault diagnosis module 602 is specifically used for:

[0144] For the cable under test, Gaussian pulse signals were injected using TDR technology at multiple detection frequencies, the same as those used in the model training phase, and the second TDR response data corresponding to each detection frequency was recorded.

[0145] Based on the second TDR response data corresponding to each detection frequency, a first detection matrix is ​​constructed;

[0146] The parameters of the normalization process during the model training phase are used to normalize the elements in the first detection matrix so that the second TDR response data at each detection frequency are on the same order of magnitude, thus obtaining the second detection matrix.

[0147] The second detection matrix is ​​input into the constructed PCA baseline model to obtain the result of whether the cable under test has a soft fault.

[0148] Furthermore, each column of the first detection matrix and / or the second detection matrix is ​​a column vector at a detection frequency, and the elements in the column vector include TDR response data of multiple sampling points corresponding to various fault scenarios characterized by fault location, fault length, and fault width.

[0149] Furthermore, the soft fault diagnosis module 602 is specifically used for:

[0150] The second detection matrix is ​​input into the constructed PCA baseline model to obtain the squared prediction error statistics for different sampling points;

[0151] If the squared prediction error statistic corresponding to multiple consecutive sampling points is greater than or equal to the fault judgment threshold, then the cable under test is determined to have a soft fault.

[0152] Furthermore, the frequency determination module 603 is used for:

[0153] For each sampling point in the fault area of ​​the cable under test where there is a soft fault, a contribution analysis is performed to calculate the contribution of each detection frequency to the squared prediction error statistic.

[0154] The detection frequency with the highest contribution is selected as the target detection frequency.

[0155] The cable soft fault diagnosis device 600 in this application embodiment can be a computer device or a component within a computer device, such as an integrated circuit or a chip. The cable soft fault diagnosis device 600 provided in this application embodiment can achieve... Figure 1 and Figure 2 The various processes implemented in the cable soft fault diagnosis method embodiment will not be described again here to avoid repetition.

[0156] This application also provides a computer device, such as... Figure 7 As shown, the computer device 700 includes a processor 701 and a memory 702. The memory 702 stores a program or instruction that can run on the processor 701. When the program or instruction is executed by the processor 701, it implements the various steps of the above-described cable soft fault diagnosis method embodiment and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0157] The memory 702 can be used to store software programs and various data. The memory 702 may primarily include a first storage area for storing programs or instructions and a second storage area for storing data. The first storage area may store the operating system, application programs or instructions required for at least one function (such as sound playback, image playback, etc.). Furthermore, the memory 702 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus RAM (DRRAM). The memory 702 in the embodiments of this application includes, but is not limited to, these and any other suitable types of memory.

[0158] Processor 701 may include one or more processing units; optionally, processor 701 integrates an application processor and a modem processor, wherein the application processor mainly handles operations involving the operating system, user interface, and applications, and the modem processor mainly handles wireless communication signals, such as a baseband processor. It is understood that the aforementioned modem processor may also not be integrated into processor 701.

[0159] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described cable soft fault diagnosis method embodiments and achieve the same technical effects. To avoid repetition, they will not be described again here.

[0160] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element. Furthermore, it should be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0161] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of this application.

Claims

1. A method for diagnosing soft faults in cables, characterized in that, include: Collect first TDR response data of fault-free cables at multiple detection frequencies, and construct a PCA benchmark model based on the first TDR response data at multiple detection frequencies; The second TDR response data of the cable under test is collected at multiple detection frequencies. Based on the second TDR response data and the established PCA benchmark model, it is determined whether there is a soft fault in the cable under test. The PCA benchmark model is used to determine whether there is a soft fault in the cable under test by calculating the squared prediction error statistic. The PCA benchmark model includes a data processing unit, a principal component extraction unit, and a fault judgment unit. The data processing unit is used to preprocess the first TDR response data. The principal component extraction unit is used to extract principal components based on the preprocessed data. The fault judgment unit is used to calculate the squared prediction error statistic based on the principal components, set the fault judgment threshold, and make a soft fault judgment based on the squared prediction error statistic and the fault judgment threshold. If a soft fault is determined in the cable under test, the contribution of each detection frequency to the squared prediction error statistic is analyzed to determine the target detection frequency.

2. The cable soft fault diagnosis method according to claim 1, characterized in that, The method involves collecting first TDR response data of fault-free cables at multiple detection frequencies, and constructing a PCA benchmark model based on this first TDR response data at multiple detection frequencies, including: For fault-free cables, Gaussian pulse signals are injected using TDR technology at multiple preset detection frequencies, and the first TDR response data corresponding to each detection frequency is recorded. A PCA benchmark model is constructed based on the first TDR response data at multiple detection frequencies.

3. The cable soft fault diagnosis method according to claim 2, characterized in that, The data processing unit is specifically used to: construct a first health matrix based on the first TDR response data corresponding to each detection frequency, and normalize the elements in the first health matrix so that the first TDR response data at each detection frequency are of the same magnitude, thereby obtaining a second health matrix; The principal component extraction unit is specifically used to: perform singular value decomposition on the second health matrix to obtain a left singular vector matrix, a singular value diagonal matrix, and a right singular vector matrix. The right singular vector matrix is ​​used to determine each principal component, which is a linear combination of the first TDR response data. The left singular vector matrix is ​​used to determine the numerical score of the principal component. The singular value diagonal matrix is ​​used to determine the variance contribution of the principal component. The unit also sorts each principal component according to its variance contribution from high to low and retains the top k principal components, where k is a positive integer. The fault judgment unit is specifically used to: calculate the squared prediction error statistic based on the retained k principal components, set a fault judgment threshold, and determine that the cable has a soft fault when the squared prediction error statistic is greater than or equal to the fault judgment threshold.

4. The cable soft fault diagnosis method according to claim 2, characterized in that, The process involves collecting second TDR response data of the cable under test at multiple identical detection frequencies. Based on this second TDR response data and the established PCA baseline model, the determination of whether the cable under test exhibits a soft fault includes: For the cable under test, Gaussian pulse signals were injected using TDR technology at multiple detection frequencies, the same as those used in the model training phase, and the second TDR response data corresponding to each detection frequency was recorded. Based on the second TDR response data corresponding to each detection frequency, a first detection matrix is ​​constructed; The parameters of the normalization process during the model training phase are used to normalize the elements in the first detection matrix so that the second TDR response data at each detection frequency are on the same order of magnitude, thus obtaining the second detection matrix. The second detection matrix is ​​input into the constructed PCA baseline model to obtain the result of whether the cable under test has a soft fault.

5. The cable soft fault diagnosis method according to claim 4, characterized in that, Each column of the first detection matrix and / or the second detection matrix is ​​a column vector at a detection frequency, and the elements in the column vector include TDR response data of multiple sampling points corresponding to various fault scenarios characterized by fault location, fault length, and fault width.

6. The cable soft fault diagnosis method according to claim 5, characterized in that, The step of inputting the second detection matrix into the constructed PCA baseline model to obtain the result of whether the cable under test has a soft fault includes: The second detection matrix is ​​input into the constructed PCA baseline model to obtain the squared prediction error statistics for different sampling points; If the squared prediction error statistic corresponding to multiple consecutive sampling points is greater than or equal to the fault judgment threshold, then the cable under test is determined to have a soft fault.

7. The cable soft fault diagnosis method according to claim 6, characterized in that, If a soft fault is determined in the cable under test, the contribution of each detection frequency to the squared prediction error statistic is analyzed to determine the target detection frequency, including: For each sampling point in the fault area of ​​the cable under test where there is a soft fault, a contribution analysis is performed to calculate the contribution of each detection frequency to the squared prediction error statistic. The detection frequency with the highest contribution is selected as the target detection frequency.

8. A cable soft fault diagnosis device, characterized in that, include: The model building module is used to collect the first TDR response data of the fault-free cable at multiple detection frequencies, and to build a PCA benchmark model based on the first TDR response data at multiple detection frequencies. The soft fault diagnosis module is used to collect the second TDR response data of the cable under test at multiple detection frequencies. Based on the second TDR response data and the established PCA benchmark model, it determines whether there is a soft fault in the cable under test. The PCA benchmark model is used to determine whether there is a soft fault in the cable under test by calculating the squared prediction error statistic. The frequency determination module is used to analyze the contribution of each detection frequency to the squared prediction error statistic if a soft fault is determined in the cable under test, in order to determine the target detection frequency.

9. A computer device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that run on the processor, the program or instructions being executed by the processor to implement the steps of the cable soft fault diagnosis method as described in any one of claims 1 to 7.

10. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or instructions are executed by the processor, they implement the steps of the cable soft fault diagnosis method as described in any one of claims 1 to 7.

Images