Mining circuit fault self-diagnosis method and system

Abstract

The invention provides a mining circuit fault self-diagnosis method and system. According to the method, current and voltage waveforms and three-dimensional vibration signals of a cable are collected, and anti-interference data are generated through self-adaptive noise reduction and time sequence synchronization; a waveform interception window is dynamically adjusted based on the correlation between the mechanical vibration intensity and the current transient rate, the wave crest slope variable quantity is extracted from the current transient segment, high-frequency harmonic energy is separated from the voltage transient segment, time-frequency analysis is carried out on the vibration signal to extract an energy sudden increase frequency point, and a mechanical damage spectrum feature set is constructed; inputting the features into a space-time correlation model, and verifying space-time consistency of current distortion and voltage abnormity to generate composite features; dynamically correcting a fault threshold based on the environmental interference factor; and finally, a short-circuit peak or open-circuit oscillation diagnosis result is output according to the association strength of the electrical and mechanical characteristics. According to the invention, accurate fault self-diagnosis of the mining cable under a complex working condition is realized.

Description

Technical Field

[0001] This application relates to the technical field of intelligent fault diagnosis and prediction of mining equipment, and particularly relates to a self-diagnosis method and system for mining circuit faults. Background Art

[0002] The underground environment of coal mines is humid, dusty and has strong electromagnetic interference. Cables are prone to high-frequency transient faults such as short-circuit spikes and open-circuit oscillations due to mechanical damage or insulation aging. Such faults have extremely short durations and weak signals. Therefore, there is an urgent need for a diagnostic system to complete acquisition, analysis and alarm within a short time, and at the same time, it is necessary to overcome the influence of complex underground working conditions on the sensor accuracy and distinguish fault signals from electromagnetic noise.

[0003] The current mainstream solution uses a high-frequency traveling wave positioning monitoring system. Fault traveling wave signals are captured by high-frequency sensors installed on the cable ground wire, and the fault distance is calculated by combining the double-end traveling wave ranging method, and the waveform data is uploaded to the cloud to trigger an alarm. The system relies on a high-speed sampling processor to achieve data acquisition and supports remote viewing of fault waveforms and positions, but only diagnoses by screening high-frequency energy through a fixed threshold.

[0004] However, this solution has significant defects. Single-signal dependence leads to the inability to distinguish short-circuit spikes from electromagnetic interference and a high false alarm rate. Feature analysis is static and only depends on the traveling wave amplitude and propagation time, making it difficult to capture the weak features of early insulation deterioration. The lack of dynamic adaptability does not integrate environmental parameters to correct the threshold, and it is prone to misjudgment during load fluctuations and the response delay exceeds the transient fault blocking requirement. Summary of the Invention

[0005] This application provides a self-diagnosis method and system for mining circuit faults to solve the problems of easy misjudgment and response delay during load fluctuations in the prior art.

[0006] In a first aspect, this application provides a self-diagnosis method for mining circuit faults, including:

[0007] Synchronously collect the current waveform, voltage waveform and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generate an anti-interference data set through adaptive noise reduction and time-sequence synchronous transmission in a humid and dusty environment;

[0008] Based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference data set, dynamically adjust the waveform intercept window to capture transient waveform segments, extract the peak slope change amount from the transient waveform segments of the current waveform, separate the harmonic frequency band energy of short-circuit spikes from the transient waveform segments of the voltage waveform, and at the same time perform time-frequency analysis on the three-dimensional vibration signal to generate a mechanical damage spectrum feature set;

[0009] Input the peak slope change amount, harmonic frequency band energy, and mechanical damage spectrum feature set into the spatio-temporal correlation model, and generate a composite feature vector through the spatio-temporal correlation verification of current distortion features and voltage anomaly features;

[0010] Based on the environmental interference factors in the composite feature vector, combined with the cable surface temperature and dust data, dynamically correct the thresholds in the fault feature database;

[0011] Perform multi-dimensional matching calculations on the composite feature vector and the corrected fault mode thresholds, and determine the fault diagnosis result according to the correlation strength between electrical features and mechanical features in the matching results.

[0012] Optionally, performing multi-dimensional matching calculations on the composite feature vector and the corrected fault mode thresholds, and determining the fault diagnosis result according to the correlation strength between electrical features and mechanical features in the matching results, includes:

[0013] Decompose the composite feature vector into an electrical feature sub-vector and a mechanical feature sub-vector, and perform Euclidean distance calculations on them respectively with the corrected fault mode thresholds to obtain the electrical matching degree and the mechanical matching degree;

[0014] Calculate the correlation strength value between the electrical matching degree and the mechanical matching degree. When the correlation strength value exceeds the preset fusion threshold, activate the fault type determination module to compare the deviation between the electrical matching degree and the short-circuit spike threshold and the deviation between the mechanical matching degree and the open-circuit oscillation threshold through the fault type determination module, and determine the fault diagnosis result according to the comparison result.

[0015] Optionally, the determining the fault diagnosis result according to the comparison result includes:

[0016] If the deviation between the electrical matching degree and the short-circuit spike threshold is less than the deviation between the mechanical matching degree and the open-circuit oscillation threshold, output the short-circuit spike diagnosis result corresponding to the cable damage;

[0017] If the deviation between the electrical matching degree and the short-circuit spike threshold is greater than the deviation between the mechanical matching degree and the open-circuit oscillation threshold, output the open-circuit oscillation diagnosis result corresponding to the electrical anomaly.

[0018] Optionally, based on the environmental interference factors in the composite feature vector, combined with the cable surface temperature and dust data, dynamically correcting the thresholds in the fault feature database includes:

[0019] Extract the environmental interference factors from the composite feature vector, and the environmental interference factors are calculated by multiplying the temperature value collected by the cable surface temperature sensor and the dust concentration monitoring value;

[0020] Select a correction factor table corresponding to the temperature compensation coefficient and the dust attenuation coefficient from the fault feature database according to the magnitude of the environmental interference factor;

[0021] Linearly combine the original threshold in the fault feature database with the temperature compensation coefficient and the dust attenuation coefficient in the correction factor table to generate a dynamically corrected threshold, where the temperature compensation coefficient has a piecewise linear relationship with the temperature value, and the dust attenuation coefficient has an exponential attenuation relationship with the dust concentration value.

[0022] Optionally, input the peak slope change amount, the harmonic frequency band energy, and the mechanical damage spectrum feature set into the spatio-temporal correlation model, and generate a composite feature vector through the spatio-temporal correlation verification of the current distortion feature and the voltage anomaly feature, including:

[0023] Input the peak slope change amount, the harmonic frequency band energy, and the mechanical damage spectrum feature set into the feature fusion module, and perform data recombination according to the time stamp alignment rule to form a spatio-temporal matrix including current distortion features, voltage anomaly features, and mechanical damage features;

[0024] Based on the current distortion feature and the voltage anomaly feature in the spatio-temporal matrix, calculate the time offset between the two within a preset time window through the cross-validation unit in the spatio-temporal correlation model. When the time offset is less than the preset threshold, trigger the space correlation unit, compare the physical position corresponding to the mechanical damage feature with the spatial coordinates of the current-voltage anomaly area, and generate a spatial coincidence degree;

[0025] Determine the correlation weight coefficients of the current distortion feature, the voltage anomaly feature, and the mechanical damage feature according to the time offset and the spatial coincidence degree, and perform weighted summation on the time offset, the spatial coincidence degree, and the correlation weight coefficients to generate a composite feature vector.

[0026] Optionally, based on the current distortion feature and the voltage anomaly feature in the spatio-temporal matrix, calculate the time offset between the two within a preset time window through the cross-validation unit in the spatio-temporal correlation model. When the time offset is less than the preset threshold, trigger the space correlation unit, compare the physical position corresponding to the mechanical damage feature with the spatial coordinates of the current-voltage anomaly area, and generate a spatial coincidence degree, including:

[0027] Extract the current distortion feature sequence and the voltage anomaly feature sequence from the spatio-temporal matrix, and perform segmented interception on the current distortion feature sequence and the voltage anomaly feature sequence according to the preset time window sliding rule to form a current distortion feature segment and a voltage anomaly feature segment containing continuous multiple sampling point feature values;

[0028] Perform point-by-point difference accumulation calculation on the current distortion feature segment and the voltage anomaly feature segment within the same time window, and minimize the phase difference between the two by dynamically adjusting the starting position of the voltage anomaly feature segment to generate the time offset within the current time window;

[0029] Determine whether the time offset is less than a preset threshold. When the time offset is less than the preset threshold, send a trigger instruction to the spatial association unit, and retrieve the physical position coordinate set corresponding to the mechanical damage feature and the spatial coordinate set of the current-voltage anomaly area through the trigger instruction;

[0030] Based on the physical position coordinate set and the spatial coordinate set of the current-voltage anomaly area, taking the minimum coverage area of the mechanical damage feature as a reference, traverse the coverage areas of all current-voltage anomaly areas, count the overlapping lengths in the horizontal and vertical ranges between the two, and generate the spatial coincidence degree according to the ratio of the product of the horizontal and vertical overlapping lengths to the area of the coverage area of the mechanical damage feature.

[0031] Optionally, based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference dataset, dynamically adjust the waveform interception window to capture transient waveform segments, including:

[0032] Based on the vibration intensity detection module of the three-dimensional vibration signal in the anti-interference dataset, calculate the root mean square value of the vibration amplitude, and establish a correlation relationship table between the vibration intensity and the current transient rate;

[0033] When it is detected that the current transient rate exceeds the rate threshold corresponding to the current vibration intensity interval in the correlation relationship table, select the slope mutation point or local extreme point within a preset time range from the current current waveform as the starting position of the waveform interception window, where the wave trough point is preferentially selected as the starting position;

[0034] Dynamically adjust the length of the waveform interception window according to the ratio of the root mean square value to the current transient rate, where the window length is extended when the ratio increases and contracted when the ratio decreases, and the window length is limited between a preset upper limit and a preset lower limit;

[0035] Analyze the amplitude mutation amplitude and the spectral energy distribution of the current waveform within the adjusted window. If the value of the amplitude mutation amplitude exceeds the preset threshold and the spectral energy is concentrated in the high-frequency band, capture the current waveform as a transient waveform segment.

[0036] Optionally, extract the peak slope change amount from the transient waveform segment of the current waveform, separate the harmonic frequency band energy of the short-circuit spike from the transient waveform segment of the voltage waveform, and simultaneously perform time-frequency analysis on the three-dimensional vibration signal to generate a mechanical damage spectral feature set, including:

[0037] From the transient waveform segments of the captured current waveform, extract the ratio of the amplitude increment to the time increment between adjacent wave peaks as the wave peak slope change amount;

[0038] Perform frequency domain decomposition on the transient waveform segments of the captured voltage waveform, identify the frequency band with a frequency higher than three times the fundamental frequency and a concentrated energy distribution, and calculate the energy integral of each sub-band within the selected frequency band as the harmonic frequency band energy;

[0039] Perform time-frequency analysis on the three-dimensional vibration signal using a multi-resolution decomposition method, extract the set of frequency points with a sudden increase in vibration energy within each decomposed frequency band, record the frequency positions and energy increases of the set of frequency points, and generate a mechanical damage spectrum feature set composed of frequency point distribution and energy change.

[0040] Optionally, synchronously collect the current waveform, voltage waveform, and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generate an anti-interference data set through adaptive noise reduction and time sequence synchronous transmission in a humid and dusty environment, including:

[0041] Collect the current waveform, voltage waveform, and three-dimensional vibration signal of the cable in a synchronous trigger mode through a sensor group composed of a current sensor, a voltage sensor, and a three-dimensional vibration sensor;

[0042] According to the signal-to-noise ratios of the current waveform and the voltage waveform, dynamically select filtering parameters through a signal intensity adaptive module, and perform high-frequency noise suppression on the current waveform and the voltage waveform based on the filtering parameters, only retaining the transient components in the signal whose amplitude fluctuations exceed a preset threshold;

[0043] When transmitting the current waveform, voltage waveform, and three-dimensional vibration signal, align the time stamps of the three axial components of the three-dimensional vibration signal, the current waveform, and the voltage waveform through a time sequence synchronous transmission module, and dynamically adjust the signal transmission interval according to the dust concentration monitoring result. When the dust concentration is higher than a preset critical value, shorten the signal transmission interval to reduce signal loss caused by dust interference;

[0044] Perform superposition analysis on the current waveform, voltage waveform, and three-dimensional vibration signal after time sequence synchronization, extract the intervals with sudden amplitude changes of the three within the same time period, and generate a set of transient signals with strictly aligned time stamps as the anti-interference data set.

[0045] In a second aspect, the present application provides a mine circuit fault self-diagnosis system, including:

[0046] An acquisition module, which synchronously collects the current waveform, voltage waveform, and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generates an anti-interference data set through adaptive noise reduction and time sequence synchronous transmission in a humid and dusty environment;

[0047] An extraction module, based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference dataset, dynamically adjusts the waveform truncation window to capture transient waveform segments, extracts the amount of change in the peak slope from the transient waveform segments of the current waveform, separates the harmonic frequency band energy of the short-circuit spike from the transient waveform segments of the voltage waveform, and simultaneously performs time-frequency analysis on the three-dimensional vibration signal to generate a mechanical damage spectrum feature set;

[0048] A correlation module inputs the amount of change in the peak slope, the harmonic frequency band energy, and the mechanical damage spectrum feature set into a spatio-temporal correlation model, and generates a composite feature vector through spatio-temporal correlation verification of the current distortion feature and the voltage anomaly feature;

[0049] A correction module dynamically corrects the threshold in the fault feature database based on the environmental interference factor in the composite feature vector, combined with the cable surface temperature and dust data;

[0050] A matching module performs multi-dimensional matching calculations on the composite feature vector and the corrected fault mode threshold, and determines the fault diagnosis result according to the correlation strength between the electrical feature and the mechanical feature in the matching result.

[0051] In the embodiment of the present application, the current waveform, voltage waveform, and three-dimensional vibration signal of the cable are synchronously collected by a multi-dimensional sensor group, and an anti-interference dataset is generated through adaptive noise reduction and time-sequence synchronous transmission in a humid and dusty environment; based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference dataset, the waveform truncation window is dynamically adjusted to capture transient waveform segments, the amount of change in the peak slope is extracted from the transient waveform segments of the current waveform, the harmonic frequency band energy of the short-circuit spike is separated from the transient waveform segments of the voltage waveform, and simultaneously time-frequency analysis is performed on the three-dimensional vibration signal to generate a mechanical damage spectrum feature set; the amount of change in the peak slope, the harmonic frequency band energy, and the mechanical damage spectrum feature set are input into a spatio-temporal correlation model, and a composite feature vector is generated through spatio-temporal correlation verification of the current distortion feature and the voltage anomaly feature; based on the environmental interference factor in the composite feature vector, combined with the cable surface temperature and dust data, the threshold in the fault feature database is dynamically corrected; the composite feature vector and the corrected fault mode threshold are subjected to multi-dimensional matching calculations, and the fault diagnosis result is determined according to the correlation strength between the electrical feature and the mechanical feature in the matching result.

[0052] The present application has the following beneficial effects:

[0053] Collect current, voltage and vibration signals through the cooperation of multi-dimensional sensors, and combine adaptive noise reduction and synchronous transmission technologies to effectively overcome the interference of the underground environment and ensure data integrity and timing consistency; adaptively intercept waveforms based on the dynamic correlation between mechanical vibration and current transient rate to accurately capture transient fault segments; extract the changes in current slope, voltage harmonic energy and vibration time-frequency characteristics, construct a multi-dimensional fault feature set, and improve the identification ability of cable insulation deterioration, mechanical damage and short-circuit faults; use the spatio-temporal correlation model to fuse the spatio-temporal synchronization of current, voltage and vibration characteristics to generate a composite feature vector, avoiding misjudgment of single signals; dynamically correct the fault threshold in combination with environmental parameters to enhance the diagnostic robustness under complex working conditions; through multi-dimensional feature matching and correlation strength analysis, achieve accurate diagnosis of high-frequency transient faults such as short-circuit spikes and open-circuit oscillations, significantly reducing the false alarm rate.

[0054] Further, by decomposing the composite feature vector into an electrical feature sub-vector and a mechanical feature sub-vector, calculating the matching degree with the corrected fault mode threshold respectively, generating the electrical matching degree and the mechanical matching degree, calculating the correlation strength value between the two and triggering the fault type determination based on the preset fusion threshold, comparing the deviation between the electrical matching degree and the short-circuit spike threshold and the deviation between the mechanical matching degree and the open-circuit oscillation threshold to determine the diagnosis result, so as to effectively distinguish the fault modes of short-circuit spikes and open-circuit oscillations, improve the diagnostic resolution of composite faults, enhance the diagnostic reliability through the dynamic correlation strength fusion and deviation comparison mechanism, and reduce the false alarm risk caused by environmental interference.

[0055] These aspects or other aspects of the present application will be more clearly understood in the following description of the embodiments. Brief Description of the Drawings

[0056] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0057] Figure 1 Shows the flowchart of a mine circuit fault self-diagnosis method provided by the present application;

[0058] Figure 2 Shows the scenario diagram of a mine circuit fault self-diagnosis method provided by the present application;

[0059] Figure 3 Shows the structural schematic diagram of a mine circuit fault self-diagnosis system provided by the present application. Detailed Description of the Embodiments

[0060] To enable those skilled in the art to better understand the solution of this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application.

[0061] In some processes described in the specification and claims of this application and the above-mentioned accompanying drawings, a number of operations appear in a specific order. However, it should be clearly understood that these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish each different operation, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions such as "first" and "second" in this article are used to distinguish different messages, devices, modules, etc., and do not represent a sequence, nor do they limit that "first" and "second" are of different types.

[0062] Researchers have found that there are significant technical bottlenecks in existing mine circuit fault diagnosis methods: single-signal dependence leads to the inability to effectively distinguish short-circuit spikes from electromagnetic interference signals, resulting in a high false alarm rate; static feature analysis only relies on traveling wave amplitude and propagation time parameters, making it difficult to capture the weak features of early insulation degradation; insufficient dynamic adaptability leads to the threshold not being corrected by environmental parameters, and it is easy to produce misjudgments and the response delay exceeds the transient fault blocking requirements in the load fluctuation scenario. Especially in the high-dust and humid environment of coal mines, traditional methods are difficult to achieve accurate extraction and dynamic verification of fault features. Therefore, there is an urgent need for an intelligent diagnosis method of multi-modal perception and dynamic correlation verification.

[0063] To address the above problems, the present invention proposes a mine circuit fault self-diagnosis method, the core of which is to construct a spatio-temporal correlation model of multi-dimensional sensing data and a dynamic threshold correction mechanism. Specifically, current waveforms, voltage waveforms, and three-dimensional vibration signals are synchronously collected by multiple sensors, and an anti-interference data set is generated by combining an adaptive noise reduction algorithm; the waveform truncation window is adjusted based on the dynamic correlation between mechanical vibration intensity and current transient rate, and the change amount of the wave peak slope, the energy of the harmonic frequency band, and the time-frequency mechanical damage characteristics are synchronously extracted; the coupling relationship between electrical characteristics and mechanical characteristics is verified through the spatio-temporal correlation model, and the threshold database is dynamically corrected in combination with environmental parameters. This method significantly reduces the false alarm rate through multi-dimensional signal fusion, enhances the detection ability of early insulation degradation by using time-frequency dynamic analysis technology, and improves the response speed under complex working conditions through the threshold correction mechanism adapted to environmental parameters, fundamentally solving the technical defects of single feature dimension, static analysis limitation, and insufficient environmental adaptability of traditional methods.

[0064] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0065] Figure 1 The figure is a flowchart of a method for self-diagnosis of mine circuit faults provided by an embodiment of the present application. As Figure 1 shown, the method includes:

[0066] 101. Synchronously collect the current waveform, voltage waveform and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generate an anti-interference data set through adaptive noise reduction and time-sequence synchronous transmission in a humid and dusty environment;

[0067] Optionally, step 101 may specifically include the following steps:

[0068] 1011. Collect the current waveform, voltage waveform and three-dimensional vibration signal of the cable in a synchronous trigger manner through a sensor group composed of a current sensor, a voltage sensor and a three-dimensional vibration sensor;

[0069] 1012. Dynamically select filtering parameters through a signal intensity adaptive module according to the signal-to-noise ratio of the current waveform and the voltage waveform, and perform high-frequency noise suppression on the current waveform and the voltage waveform based on the filtering parameters, and only retain the transient components in the signal whose amplitude fluctuations exceed a preset threshold;

[0070] 1013. When transmitting the current waveform, voltage waveform and three-dimensional vibration signal, align the time stamps of the three axial components of the three-dimensional vibration signal, the current waveform and the voltage waveform through a time-sequence synchronous transmission module, and dynamically adjust the signal transmission interval according to the dust concentration monitoring result. When the dust concentration is higher than a preset critical value, shorten the signal transmission interval to reduce signal loss caused by dust interference;

[0071] 1014. Perform superposition analysis on the current waveform, voltage waveform and three-dimensional vibration signal after time-sequence synchronization, extract the intervals where the amplitudes mutate in the same time period of the three, and generate a set of transient signals with strictly aligned time stamps as the anti-interference data set.

[0072] In the above steps, the multi-dimensional sensor group refers to a device combination composed of a current sensor, a voltage sensor, and a three-dimensional vibration sensor. Adaptive noise reduction refers to a technology that dynamically adjusts filtering parameters according to the signal-to-noise ratio of the current waveform and the voltage waveform to suppress high-frequency noise. The timing synchronization transmission module refers to a device that aligns the timestamps of the three axial components of the three-dimensional vibration signal, the current waveform, and the voltage waveform. The anti-interference data set refers to a set of transient signals generated by extracting the intervals of amplitude mutations of the current waveform, the voltage waveform, and the three-dimensional vibration signal in the same time period through superposition analysis. The transient component refers to the short-term changing part of the signal whose amplitude fluctuation exceeds a preset threshold. The dust concentration monitoring result refers to the dynamic data of the environmental dust content obtained by the dust sensor.

[0073] In the embodiment of the present application, first, a set of combined sensors (current sensor, voltage sensor, and three-dimensional vibration sensor) synchronously collect key signals of the cable. Specifically, the synchronous signal collection is realized through step 1011. The current sensor uses the Hall effect principle to collect the cable current waveform, the voltage sensor obtains the voltage waveform based on the voltage division circuit, and the three-dimensional vibration sensor synchronously measures the vibration accelerations of the X / Y / Z three axes through piezoelectric ceramic elements. The three groups of sensors are synchronously started through the hardware trigger circuit, and the main control unit sends a unified trigger pulse signal to ensure that the time starting points of the current waveform, the voltage waveform, and the three-dimensional vibration signal are aligned, and eliminate the timing deviation caused by the sensor response delay. In order to ensure that these different signals are completely aligned in time (to avoid data misalignment caused by different sensor response speeds), they do not start working independently, but a main controller sends a unified "start" electrical pulse signal to trigger the three sensors to start collecting data at the same time, just like pressing the start keys of multiple stopwatches at the same time, ensuring that the time starting points of all signals are the same.

[0074] Since the collected original current and voltage waveform signals will inevitably be mixed with various high-frequency interference noises. Secondly, the adaptive noise reduction process is completed through step 1012. The signal strength adaptive module (wherein, the signal strength adaptive module is implemented by a software program embedded in the main control unit, and the code logic in the adaptive filter algorithm library is run in real time by the processor. Its core function is to automatically adjust the subsequent filter parameters according to the signal quality index signal-to-noise ratio calculated in real time.) calculates the signal-to-noise ratio of the current waveform and voltage waveform using the sliding window method, and updates the signal-to-noise ratio value every 0.1 seconds. When the detected signal-to-noise ratio is lower than 20 dB, a Butterworth low-pass filter is automatically selected and the cut-off frequency is set to 500 Hz; when the signal-to-noise ratio is higher than 40 dB, it switches to a Chebyshev filter and the cut-off frequency is increased to 2 kHz. Based on the dynamically selected filter parameters, digital filtering is performed on the original signal, and only the transient components with amplitude fluctuations exceeding 2% of the baseline value are retained, while power frequency interference and random high-frequency noises are eliminated. That is to say, the system will automatically adjust the noise reduction intensity according to the quality of the signal itself (signal-to-noise ratio). The system continuously (every 0.1 seconds) calculates the signal-to-noise ratio of a small section of the current or voltage waveform (measuring the ratio of the useful part to the noise part in the signal). The higher the signal-to-noise ratio, the better the signal quality; the lower the signal-to-noise ratio, the more serious the noise pollution. The system presets the judgment value of the signal-to-noise ratio: when the calculated signal-to-noise ratio is lower than 20 decibels (dB, the unit of sound or signal strength, the lower the value, the greater the noise), it means the noise is large, and the system will automatically select a Butterworth low-pass filter and set the filter to only allow signal components with frequencies lower than 500 Hertz (Hz) to pass through (i.e., retain the useful signals with lower frequencies and filter out the high-frequency noises); on the contrary, when the signal-to-noise ratio is higher than 40 dB, it means the signal quality is good, and the system will switch to a Chebyshev low-pass filter and increase the upper limit of the signal frequency allowed to pass through to 2000 Hertz (2 kHz), so that more useful high-frequency details can be retained. Regardless of which filter and setting are used, the ultimate goal is to eliminate those subtle and useless high-frequency fluctuations (power frequency interference and random noise), and only retain those short abnormal signals (transient components) with relatively significant amplitude changes (fluctuation amplitude exceeding 2% of the normal baseline value).

[0075] Next, the processed current, voltage signals, and three-dimensional vibration signals (including data in the X, Y, and Z directions) need to be transmitted to the data processing center. In a humid and dusty environment, transmission signals are prone to interference and even loss. To solve this problem and ensure that all signals are precisely synchronized in time, the system adopts a timing synchronization transmission technology. Specifically, the timing synchronization transmission is achieved through step 1013. The timing synchronization transmission module uses the Network Time Protocol to add nanosecond-level timestamps to the X / Y / Z components of the three-dimensional vibration signal, the filtered current waveform, and the voltage waveform. The time deviation of each signal channel is identified and corrected through an abnormal timestamp detection algorithm to ensure that the time alignment error of multiple signals is less than 1 ms. The dust concentration monitoring unit collects dust data using the light scattering principle. When the detected dust concentration exceeds 200 mg / m³, it triggers a transmission interval adjustment algorithm to dynamically shorten the signal transmission interval from the default 100 ms to 50 ms, reducing the signal loss rate caused by dust by increasing the data packet sending frequency. That is to say, the system will use a high-precision time protocol (such as a trimmed-down version of the Network Time Protocol NTP or a hardware timestamp) to attach an accurate time tag (timestamp) accurate to the nanosecond (one billionth of a second) level to each collected data point (including current values, voltage values, X / Y / Z vibration values). An algorithm will also run inside the system to specifically check whether the time tags on different signal channels are strictly aligned. If a slight deviation (such as being 0.5 milliseconds slower) is found in the time tag of a certain signal, the algorithm will automatically correct it to ensure that the time error between all signals is less than 1 millisecond. At the same time, the dust concentration in the environment is continuously monitored by a dust sensor (usually by emitting a beam of light and measuring the intensity of the light scattered back by dust particles to calculate the concentration). This dust concentration data is used to dynamically adjust the speed of signal transmission: when the dust concentration is normal (below 200 milligrams per cubic meter - mg / m³), the system defaults to sending data packets every 100 milliseconds (ms); when the dust sensor detects that the concentration has risen sharply above 200 mg / m³ (the preset critical value), the system will immediately initiate an adjustment to shorten the interval between data packet transmissions to 50 milliseconds each time.

[0076] Finally, all the current, voltage, and X / Y / Z three-dimensional vibration signals that have been precisely time-aligned (with consistent timestamps) and noise-reduced are gathered together for comprehensive analysis, and an anti-interference dataset is generated through step 1014. Perform time-domain superposition analysis on the current waveform, voltage waveform, and three-dimensional vibration signal after time synchronization. Use the sliding window algorithm to traverse all signal data with a step size of 10 ms. Detect the amplitude mutation intervals where all three exceed the baseline value by 30% within the same time window through the threshold comparison module, and use wavelet transform to extract the time-frequency characteristics of each signal segment. Reorganize the selected transient signal segments according to the nanosecond-level timestamps to form a multi-dimensional anti-interference dataset containing waveform amplitudes, vibration vectors, and environmental dust concentrations, providing high-precision input for subsequent fault diagnosis. That is to say, the system sets a "time window" with a width of 10 milliseconds. Starting from the time point when the signal begins, this window slides over the data of the entire time period in steps of 10 milliseconds. Within each 10-millisecond small window, the system will simultaneously check the values of the current, voltage, and three-dimensional vibration (usually looking at the combined vibration intensity in three directions). Only when the amplitudes of all three signals show a significant abnormal increase (the amplitude changes exceed 30% of their respective normal baseline values) within this small window, the system considers this to be a "transient event" worthy of attention caused by some potential problems (such as short circuits, mechanical shocks). For example, at a certain time point, the current suddenly surges from 100 A to 150 A (exceeding 30% of the baseline 100 A, which is 130 A), while the voltage drops suddenly from 220 V to 150 V (lower than 30% of the baseline 220 V, which is 154 V), and the vibration intensity also suddenly increases by more than 30% of the normal level. When all three occur within a 10-millisecond window, they will be captured. For these captured transient event intervals, the system will use wavelet transform technology (a tool for simultaneously analyzing the time and frequency characteristics of signals) to extract more detailed characteristic information of the signal waveform. Finally, the system reorganizes and packages all the selected transient event segments, together with their precise nanosecond-level time tags, corresponding waveform amplitudes, three-dimensional vibration direction data, and the dust concentration values at that time, to form a multi-dimensional, strictly time-aligned, and anomaly-emphasized "anti-interference dataset". This high-quality dataset provides a reliable basis for accurately judging whether there are faults in the cable and what the faults are in the subsequent process.

[0077] In practical applications, for example, in a coal mine main transportation lane cable intelligent monitoring project, a collection system consisting of HX-10C high-frequency current sensor, PT-8K voltage sensor and three-axis MEMS vibration array was deployed, and the GPS synchronous clock trigger mechanism was used to achieve 5ms time synchronization. The system continuously collects the three-phase current, ground voltage and X / Y / Z axial vibration signals of the cable at a sampling rate of 50kHz. When the signal-to-noise ratio of the A-phase current is detected to drop to 25dB, the adaptive module switches to a 6th-order Chebyshev filter, sets the passband cutoff frequency to 12kHz and the stopband attenuation to 60dB, filters out the high-frequency harmonics generated by the start and stop of the tunnel boring machine, and only retains transient current fluctuations that last more than 3ms and have an amplitude exceeding 20% of the rated value. When the environmental monitoring unit feedbacks that the lane humidity reaches 95%RH, the timing synchronization module starts the redundant check mechanism, improves the time alignment accuracy of the three-axis vibration signal and the electrical signal to ±0.2ms, and reduces the data packet retransmission rate under dust interference to 1 / 3 of the industry standard by adding CRC check bits. The data analysis platform performed correlation mining on 3.6TB of raw data sampled continuously for 8 hours, identified the spatiotemporal coupling events of 132-order current harmonic distortion and the Z-axis 5-8kHz vibration spectrum mutation, and constructed an anti-interference database containing millisecond-level fault feature fragments, providing high-confidence samples for subsequent insulation layer damage prediction.

[0078] In the overall solution of the above step 101, the current and voltage waveforms and three-dimensional vibration signals of the cable during operation are collaboratively collected through a multi-dimensional sensor group, and the signal strength adaptive module is used to dynamically match the filter parameters to achieve high-frequency noise suppression. The time stamps of multi-source signals are accurately aligned in combination with the timing synchronous transmission mechanism, and the transmission interval is dynamically compressed based on the dust concentration to reduce the signal loss rate. The synchronous transient feature interval is extracted by multi-modal signal superposition analysis to construct a full-time domain fault feature data set with strong anti-interference ability, effectively overcome the dust and moisture interference and electromagnetic noise coupling problems under complex working conditions in mines, and significantly improve the recognition accuracy and integrity of cable fault characteristics. At the same time, the adaptive transmission strategy is used to ensure the accuracy and stability of signal acquisition, provide high-reliability data support for early warning and precise positioning of faults in underground circuit systems, greatly reduce the risk of misjudgment and missed detection due to environmental interference, and enhance the active safety protection capability and operation and maintenance efficiency of mine power supply systems.

[0079] 102. Based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference data set, dynamically adjust the waveform capture window to capture the transient waveform segment, extract the peak slope change from the transient waveform segment of the current waveform, separate the harmonic frequency band energy of the short-circuit spike from the transient waveform segment of the voltage waveform, and simultaneously perform time-frequency analysis on the three-dimensional vibration signal to generate a mechanical damage spectrum feature set;

[0080] Optionally, step 102 may include the following specific steps:

[0081] 1021. A vibration intensity detection module based on the three-dimensional vibration signals in the anti-interference dataset calculates the root mean square value of the vibration amplitude and establishes a correlation table between the vibration intensity and the current transient rate.

[0082] 1022. When it is detected that the current transient rate exceeds the rate threshold corresponding to the current vibration intensity range in the correlation table, a slope mutation point or a local extreme point within a preset time range is selected from the current current waveform as the starting position of the waveform truncation window, where the wave valley point is preferentially selected as the starting position.

[0083] 1023. According to the ratio of the root mean square value to the current transient rate, the length of the waveform truncation window is dynamically adjusted, where the window length is extended when the ratio increases and contracted when the ratio decreases, and the window length is restricted between a preset upper limit and a preset lower limit.

[0084] 1024. Analyze the amplitude mutation amplitude and the spectral energy distribution of the current waveform within the adjusted window. If the value of the amplitude mutation amplitude exceeds a preset threshold and the spectral energy is concentrated in the high-frequency band, the current waveform is captured as a transient waveform segment.

[0085] 1025. From the transient waveform segment of the captured current waveform, the ratio of the amplitude increment between adjacent wave peaks to the time increment is extracted as the wave peak slope change amount.

[0086] 1026. Perform frequency-domain decomposition on the transient waveform segment of the captured voltage waveform, identify the frequency band with a frequency higher than three times the fundamental frequency and a concentrated energy distribution, and calculate the energy integral of each sub-band within the selected frequency band as the harmonic band energy.

[0087] 1027. Perform time-frequency analysis on the three-dimensional vibration signal using a multi-resolution decomposition method, extract the set of frequency points with a sudden increase in vibration energy within each decomposition frequency band, record the frequency positions and energy increase amplitudes of the set of frequency points, and generate a mechanical damage spectral feature set composed of the frequency point distribution and the energy change.

[0088] In the above steps, the mechanical vibration intensity refers to the vibration energy level quantified by the root mean square value of the three-dimensional vibration signal. The current transient rate refers to the number of mutations in which the current waveform amplitude exceeds the preset threshold per unit time. The vibration intensity detection module refers to the processing unit that calculates the root mean square value of the three-dimensional vibration signal. The association table refers to a two-dimensional data table that stores the current transient rate thresholds corresponding to different vibration intensity intervals. The waveform interception window refers to the time range for capturing transient waveforms defined according to dynamic rules. The trough point refers to the lowest amplitude position between adjacent peaks in the current waveform. The local extreme point refers to the sampling point where the amplitude reaches the local maximum or minimum value in the waveform. The peak slope change refers to the ratio of the adjacent peak amplitude increment to the time increment in the current transient segment. Frequency domain decomposition refers to the process of converting the time domain signal into the frequency domain energy distribution. The harmonic frequency band energy refers to the energy integral value of the frequency band with an integer multiple of the fundamental frequency in the voltage transient segment. The multi-resolution decomposition method refers to the analysis technology of decomposing the signal into different time-frequency resolution layers by wavelet packet transform. The energy sudden increase frequency point refers to the frequency domain position where the energy increase exceeds 40% at a specific frequency of the vibration signal. The mechanical damage spectrum feature set refers to the set matrix containing the damage characteristic frequency and its energy increase parameter.

[0089] In the embodiment of the present application, it is found that there is a certain correlation law between the mechanical vibration intensity of the cable (representing the physical intensity of the shaking) and the transient change rate of the current signal (representing the frequency of abnormal changes in the current in a short period of time) (this law forms a correlation table through data accumulation). The system will use this law to intelligently "focus" on observation and dynamically adjust the time window of analysis (that is, which small time period to focus on) to more accurately capture those transient signal fragments that can reflect the problem. First, a correlation table is constructed through step 1021. The vibration intensity detection module adopts a moving root mean square algorithm to calculate the root mean square value of the three-dimensional vibration signal X / Y / Z three-axis composite vector with a window of 100ms, and synchronously counts the number of mutations of the current waveform whose amplitude exceeds 15% of the baseline value every 5 seconds as the transient rate. The root mean square value of the vibration intensity is divided into 20 intensity intervals at an interval of 0.5g, and the maximum transient rate threshold allowed in each interval is recorded to form an association table with intensity interval numbers and corresponding speed limit values. That is to say, the system will calculate the comprehensive vibration intensity of the three-dimensional vibration signal (X / Y / Z directions) using the mathematical method of root mean square (RMS). At the same time, by analyzing a large amount of historical data, the system divides the vibration intensity into 20 levels (intensity intervals) at intervals of 0.5g (unit of gravity acceleration), and sets a maximum allowable current transient rate threshold for each vibration intensity level (for example, when the vibration intensity is between 1.0g and 1.5g, if the current transient rate exceeds 3 times per second, there may be a problem), thus establishing an association table.

[0090] Secondly, the system will monitor in real time whether the current transient rate exceeds the safety threshold corresponding to the current vibration intensity. Once the current transient rate is found to be excessive (for example, the actual rate reaches 3.5 times per second, exceeding the 2.8 times / second corresponding to 1.2g), the system will consider that an abnormal event that needs to be focused on is likely to have occurred at this time. The starting position of the window is determined by step 1022. When the current transient rate exceeds the associated threshold corresponding to the current vibration intensity, the current waveform is subjected to extreme value detection. The first-order derivative between adjacent sampling points is calculated by the sliding difference method, and when the derivative sign is detected to change from negative to positive, it is marked as a trough point. If no trough point is found for 10 consecutive sampling cycles, the local extreme value detection algorithm is used to find the sampling point with the largest amplitude change rate as the starting position of the window to ensure that the interception window covers the initial stage of the mutation. In other words, in order to capture the key signal fragment (transient waveform fragment) of this event, the system needs to determine at which time point of the current waveform to start "focusing observation". The system preferentially looks for the "trough point" (the lowest point between two adjacent peaks) on the current waveform, which is usually the starting point of an abnormal event (just like there may be a small drop before the arrival of the earthquake wave). The system identifies the current by calculating the rate of change (slope) between adjacent points on the current waveform: the turning point where the rate of change suddenly changes from a negative number (indicating a decrease) to a positive number (indicating the beginning of an increase) is marked as a trough point. If no such trough point is found within 10 consecutive sampling points (representing a very short period of time, such as 0.1 milliseconds), the system will turn to look for the point on the current waveform with the most dramatic change (the absolute value of the rate of change is the largest) as the starting point for observation.

[0091] Next, after determining the starting point of the observation, the system also needs to decide how long this "focused observation" time window should last. The window length is adjusted through step 1023. According to the ratio of the root mean square value of the vibration intensity to the transient rate of the current, a linear interpolation algorithm is used to dynamically adjust the window length within the preset range of 50 - 300 ms. It is set that when the ratio increases by 0.1, the window expands by 10 ms, and when the dust concentration exceeds 200 mg / m³, the length compensation mechanism is activated to expand the window by an additional 5 ms. At the same time, the upper limit of the window length is set not to exceed the maximum duration of the fault feature, which is 300 ms, and the lower limit is not less than the minimum effective feature duration of 50 ms. That is to say, the window length is not fixed, but is dynamically adjusted according to the current ratio of the root mean square value of the vibration intensity to the transient rate of the current. This ratio can be understood as "the vibration intensity corresponding to the unit current mutation rate". If this ratio becomes larger (indicating that the vibration is strong but the current mutation is relatively small, which may imply a problem mainly caused by mechanical shock), the system will appropriately extend the observation window (each time the ratio increases by 0.1, the window extends by 10 milliseconds) to see more comprehensively and capture a possibly longer vibration process. On the contrary, if the ratio becomes smaller (indicating that the current mutation is very frequent but the vibration is relatively weak, which may imply an electrical problem), the system will shorten the observation window to focus more on the short current changes. At the same time, considering the limitation of the actual fault feature duration, the window length is strictly limited between 50 milliseconds (the shortest) and 300 milliseconds (the longest). Additionally, if the environmental dust concentration is very high (exceeding 200 mg / m³), the system will additionally extend the window by 5 ms to compensate for the loss of signal details that may be caused by dust interference.

[0092] Then, within the dynamically adjusted time window, the system will carefully examine the current waveform to determine whether it truly contains valuable transient event segments. Transient segments are captured through step 1024. The current waveform within the window is subjected to a fast Fourier transform to calculate the energy proportion in the high-frequency band of 500 - 2000 Hz. The maximum amplitude mutation within the synchronous detection window is detected. When the high-frequency energy proportion exceeds 60% and the amplitude mutation reaches 25% of the baseline value, a capture instruction is triggered for the dual-threshold comparator output. The current and voltage waveforms within the qualified time window are marked as transient segments, and their start timestamps are recorded. That is to say, it mainly looks at two points: one is whether the maximum instantaneous change amplitude of the current within the window is large enough (exceeding 25% of the baseline value); the other is whether the current signal within this window contains rich high-frequency components (whether the proportion of the energy in the range of 500 Hz to 2000 Hz in the total signal energy exceeds 60%). Only when both of these conditions are met will the system confirm that the captured current and voltage signals within the window are valid transient waveform segments. For example, if the normal current baseline is 100 A, the current within the window suddenly surges above 125 A (exceeding 25%), and the proportion of the energy of the high-frequency hiss in the current signal within this window is high (>60%), then the capture is successful.

[0093] Subsequently, for the captured current transient segments, the system will further extract a key feature: the change amount of the peak slope. The peak slope is calculated through step 1025. Cubic spline interpolation processing is performed on the captured current transient segments, and the adjacent peak positions are identified using the adaptive threshold method. The amplitude difference and time difference of each pair of peaks in the first three complete peak cycles are calculated, and the median of the three is taken as the change amount of the peak slope. The area compensation algorithm is used to eliminate measurement errors for asymmetric waveforms, and four significant figures after the decimal point are retained. That is to say, this describes how fast the current climbs from one peak to the next. The system will first accurately find the positions of several consecutive peaks within the segment (using the smoothing interpolation and adaptive threshold methods). Then, within the first three complete peak cycles, the height difference (amplitude increment) and time difference (time increment) between each adjacent pair of peaks are calculated, and the ratio of these two increments (slope = height difference / time difference) is obtained. The median of these three slope values (the middle value after sorting) is taken as the final change amount of the peak slope, which can avoid the influence of individual outliers.

[0094] After that, for the captured voltage transient segments, the system focuses on the "short-circuit spike" features therein, especially those high-frequency harmonic components. The harmonic energy is extracted through step 1026. The voltage transient segments are subjected to six-layer wavelet packet decomposition to reconstruct 32 sub-bands with a fundamental frequency three times or higher (150 - 1000 Hz). The improved Sine function is used to calculate the energy integral value of each sub-band within a 20 ms time window, and the characteristic frequency bands with integral values exceeding three times the background noise level are screened. The energy integral values of the selected frequency bands are summed up as the harmonic frequency band energy eigenvalue of the transient segment. That is to say, the system decomposes the voltage signal into different frequency bands (frequency-domain decomposition), focusing on the frequency bands with a frequency higher than three times the fundamental frequency (usually 50 Hz or 60 Hz), i.e., above 150 Hz. Six-layer wavelet packet decomposition is performed to divide the range from 150 Hz to 1000 Hz into 32 finer sub-bands. For each sub-band, the system calculates its energy integral value within a 20 ms time window. Then, the system only selects those characteristic sub-bands with energy more than three times higher than the normal background noise and relatively concentrated energy. By adding up the energy integral values of these selected sub-bands, the "harmonic frequency band energy" representing the short-circuit spike features is obtained.

[0095] Finally, for the three-dimensional vibration signals (X / Y / Z), the system performs a more refined time-frequency analysis (looking at both the variation of vibration intensity over time and its distribution at different frequencies) with the aim of extracting unique "fingerprint" spectral features that can characterize mechanical damage (such as insulation wear, structural looseness). A feature set is generated through step 1027. The three-dimensional vibration signal is subjected to eight-layer wavelet packet decomposition, and the Hilbert-Huang marginal spectrum is calculated in 24 sub-bands at each layer. The sliding energy detection method is used to identify the frequency points in each sub-band where the energy increase exceeds 40%, and the central frequency value and the percentage increase are recorded. All the damage feature frequency points are classified at 50Hz intervals to form a matrix of mechanical damage spectral feature sets that includes frequency distribution, increase intensity, and duration. That is to say, the system uses the eight-layer wavelet packet decomposition method to break down the vibration signal into many components with different frequency resolutions. In each of the decomposed sub-bands, the system calculates the marginal spectrum of the signal (a spectrum that can reflect the energy strength at each frequency point), and uses the sliding window method to detect which specific frequency points (frequency points) have a sudden significant increase in vibration energy within a short period of time (the increase exceeds 40%). The system records the specific frequency position of these energy surge frequency points and how much the energy has increased (percentage increase). Finally, all the detected such feature frequency points are classified and sorted at 50Hz intervals (for example, grouping all the damage feature points between 50-100Hz, 100-150Hz...) to form a detailed "mechanical damage spectral feature set". This feature set records at which frequency points abnormal intense vibrations are found (which may correspond to specific types of mechanical damage) and the degree of these vibration abnormalities.

[0096] In practical applications, for example, in a certain coal mine intelligent cable monitoring system, when the shearer cutting gangue causes cable anomalies, the system tracks the vibration signal at a sampling rate of 2000Hz and detects that the root mean square value of the X-axis vibration suddenly increases from the daily 1.2g to 3.8g. At the same time, the associated current transient rate breaks through 120 threshold. At this time, the analysis module locks the deep trough point in the current waveform, such as the 280A instantaneous depression point that appears at a reference current of 380A, as the starting mark of the waveform interception window. According to the dynamic ratio of the vibration intensity to the current change rate (such as when 3.8g vibration corresponds to 0.85g / ( ), the window length is automatically extended from the default 20ms to 35ms to fully capture the current waveform segment containing three consecutive peak mutations. For the intercepted current transient segment, the system accurately calculates the adjacent peak features: for example, the first to the second peak jumps from 420A to 680A within 2ms, generating a slope mutation of 130 ; the third peak decays to 510A within the subsequent 3ms, forming a -56.7 Negative rate of change. After the voltage waveform segment for synchronous analysis undergoes a 2048-point FFT transformation, an energy concentration phenomenon is detected in the high-frequency region of 1500 - 4500 Hz. The energy integral value of the 2350 Hz frequency band reaches 17 times that of the standard working condition, clearly indicating the characteristics of short-circuit discharge. The three-dimensional vibration signal is decomposed by five-layer wavelet packet decomposition, and an energy sudden increase event is captured in the third decomposition layer (frequency band of 560 - 1200 Hz): the vibration energy of the Z-axis surges from the normal state of 5 m / s² to 22 m / s² at 850 Hz, and at the same time, the energy of the X-axis 720 Hz component increases by 9 times, forming a damage fingerprint map containing frequency point coordinates, duration, and energy gradient. These characteristic data are mapped in the three-dimensional model of the roadway to help engineers accurately locate the mechanical damage risk of the insulation layer of the cable section near the No. E15 support.

[0097] In the overall scheme of step 102 above, the position and length of the waveform intercept window are adaptively adjusted through the dynamic correlation model of mechanical vibration intensity and current transient rate. The wave trough point is preferentially selected as the starting position, and the window boundary is elastically expanded or contracted according to the ratio to accurately capture the transient waveform segment with amplitude mutation and high-frequency energy concentration. Combining the change amount of the wave peak slope extraction and the harmonic frequency band energy integral calculation, the multi-resolution decomposition method is simultaneously used to perform time-frequency analysis on the vibration signal to locate the frequency point of energy sudden increase, and a composite fault characterization system integrating electrical transient characteristics and mechanical damage spectrum is constructed, effectively solving the problem of missed detection of transient characteristics caused by fixed window interception in traditional methods, strengthening the correlation mining ability of weak short-circuit spikes and vibration energy mutation, significantly improving the coupling diagnosis accuracy of early mechanical damage and electrical faults of the cable, and at the same time enhancing the interpretability of fault types through the quantitative description of frequency point distribution and energy increase, providing multi-dimensional feature support for the accurate identification and fault tracing of hidden cable defects under complex working conditions.

[0098] 103. Input the wave peak slope change amount, harmonic frequency band energy, and mechanical damage spectrum feature set into the spatio-temporal correlation model, and generate a composite feature vector through the spatio-temporal correlation verification of current distortion characteristics and voltage anomaly characteristics;

[0099] Optionally, step 103 may specifically include the following steps:

[0100] 1031. Input the wave peak slope change amount, harmonic frequency band energy, and mechanical damage spectrum feature set into the feature fusion module, and perform data reorganization according to the time stamp alignment rule to form a spatio-temporal matrix containing current distortion characteristics, voltage anomaly characteristics, and mechanical damage characteristics;

[0101] 1032. Based on the current distortion characteristics and voltage anomaly characteristics in the spatio-temporal matrix, calculate the time offset between the two within a preset time window through the cross-validation unit in the spatio-temporal correlation model. When the time offset is less than the preset threshold, trigger the spatial correlation unit, compare the physical position corresponding to the mechanical damage characteristics with the spatial coordinates of the current-voltage anomaly area, and generate the spatial coincidence degree.

[0102] Among them, step 1032 may specifically include the following process: Extract the current distortion feature sequence and voltage anomaly feature sequence from the spatio-temporal matrix, segment and intercept the current distortion feature sequence and voltage anomaly feature sequence according to the preset time window sliding rule to form a current distortion feature segment and a voltage anomaly feature segment containing consecutive multiple sampling point feature values; perform point-by-point difference accumulation calculation on the current distortion feature segment and voltage anomaly feature segment within the same time window, and minimize the phase difference between the two by dynamically adjusting the starting position of the voltage anomaly feature segment to generate the time offset within the current time window; determine whether the time offset is less than the preset threshold. When the time offset is less than the preset threshold, send a trigger instruction to the spatial correlation unit, and retrieve the physical position coordinate set corresponding to the mechanical damage characteristics and the spatial coordinate set of the current-voltage anomaly area through the trigger instruction; based on the physical position coordinate set and the spatial coordinate set of the current-voltage anomaly area, taking the minimum coverage area of the mechanical damage characteristics as the benchmark, traverse the coverage areas of all current-voltage anomaly areas, count the overlapping lengths in the horizontal and vertical ranges of the two, and generate the spatial coincidence degree according to the ratio of the product of the horizontal and vertical overlapping lengths to the area of the coverage area of the mechanical damage characteristics.

[0103] 1033. Determine the correlation weight coefficients of the current distortion characteristics, voltage anomaly characteristics, and mechanical damage characteristics according to the time offset and spatial coincidence degree, and perform weighted summation on the time offset, spatial coincidence degree, and the correlation weight coefficients to generate a composite feature vector.

[0104] In the above steps, the spatio-temporal correlation model refers to an analysis model that fuses the spatio-temporal relationships of current distortion characteristics, voltage anomaly characteristics, and mechanical damage characteristics. The feature fusion module refers to a processor that reorganizes multi-dimensional features according to the time stamp alignment rule. The spatio-temporal matrix refers to a three-dimensional data structure containing time dimension, space dimension, and feature dimension. The cross-validation unit refers to a processing unit that verifies the time series relationship between current and voltage characteristics. The time offset refers to the absolute value of the time difference when the current distortion characteristics and voltage anomaly characteristics appear. The spatial correlation unit refers to an analysis module that verifies the spatial position relationship between mechanical damage and electrical anomalies. The spatial coincidence degree refers to the overlapping ratio of the mechanical damage area and the electrical anomaly area in spatial coordinates. The correlation weight coefficient refers to the feature importance coefficient calculated according to the spatio-temporal matching degree. The composite feature vector refers to a multi-dimensional feature set after weighted fusion of spatio-temporal features.

[0105] In the embodiments of the present application, the key features extracted in the previous steps (the change amount of the current peak slope, the energy of the voltage harmonic frequency band, and the mechanical damage spectrum feature set) are comprehensively analyzed to verify their correlation in time and space, and finally fused to generate a "composite feature vector" that can more comprehensively reflect the abnormal state of the cable. The features of only a single signal are not reliable enough. It is necessary to check whether the electrical abnormal features of the current and voltage occur almost simultaneously in time (temporal correlation), and whether they coincide with the mechanical damage position detected by vibration in space (spatial correlation). This double verification in time and space can significantly improve the accuracy of fault judgment. First, a spatio-temporal matrix is constructed through step 1031. The feature fusion module uses a timestamp matching algorithm to align the change amount of the peak slope, the energy of the harmonic frequency band, and the mechanical damage spectrum feature set according to millisecond-level timestamps. Cubic spline interpolation is used to compensate for the feature data with missing timestamps, forming a three-dimensional matrix including a time axis (0 - 1000 ms), a space axis (cable coordinates X / Y / Z), and a feature axis (current / voltage / vibration features), where each data unit stores the feature value corresponding to the spatio-temporal point. That is to say, first, the system inputs the feature data from these three sources (the change amount of the current peak slope, the energy of the voltage harmonic frequency band, and the mechanical damage spectrum feature set) into a feature fusion module. The core task of this module is to reorganize and align these data according to accurate timestamps (millisecond-level) to ensure that the current features, voltage features, and vibration features at the same time point are placed together. For the feature data missing at very few time points, the system will use a mathematical interpolation method (cubic spline interpolation) for reasonable estimation and filling. Finally, all these feature data aligned in time, together with their corresponding spatial position information (at which point on the cable the current and voltage are abnormal and at which position mechanical damage is detected), are organized into a structured three-dimensional data table - the spatio-temporal matrix. The three dimensions of this matrix are: the time axis (for example, from 0 ms to 1000 ms), the space axis (the specific position coordinates of the cable, such as X / Y / Z or a length mark along the cable), and the feature axis (including the current distortion feature value, the voltage abnormal feature value, and the mechanical damage feature value).

[0106] Secondly, the system uses the constructed spatio-temporal matrix to verify the synchronism in time between the current distortion feature and the voltage abnormal feature through the internal "spatio-temporal correlation model". The spatio-temporal correlation model refers to an analysis model that fuses the spatio-temporal relationships of the current distortion feature, the voltage abnormal feature, and the mechanical damage feature. The spatio-temporal correlation verification is performed through step 1032. The cross-verification unit performs cross-correlation analysis on the current distortion feature sequence and the voltage abnormal feature sequence in the spatio-temporal matrix, and intercepts the feature segments with a sliding window step size of 10 ms. The phase alignment algorithm is used to dynamically adjust the starting position of the voltage feature segment, and the cumulative value of the difference between the two feature segments is minimized Determine the optimal time offset. When the detected offset is less than 5 ms, the spatial correlation unit activates the position matching algorithm, and performs a spatial projection comparison between the three-dimensional coordinate set located by the time difference of vibration signal propagation corresponding to the mechanical damage feature and the coordinates of the current-voltage abnormal area located by the impedance method, and calculates the geometric mean of the horizontal overlap rate (Lx / Ltotal) and the vertical overlap rate (Ly / Ltotal) as the spatial coincidence degree. That is to say, the "cross-verification unit" extracts the sequence of current distortion feature values changing with time (current distortion feature sequence) and the sequence of voltage abnormal feature values changing with time (voltage abnormal feature sequence) from the matrix. Then, it sets a sliding time window (for example, 100 milliseconds wide) and lets this window slide step by step along the time axis (for example, each time it slides 10 milliseconds). Within each window position, it extracts a small segment of the current feature sequence and a small segment of the voltage feature sequence. Then, the system attempts to finely adjust the starting position of the voltage feature sequence forward and backward within this window, calculates the absolute value of the difference between the feature values of the current sequence and the (shifted) voltage sequence at each corresponding time point, and accumulates these difference values. The system searches for the starting position offset (Δt) of the voltage sequence that makes this accumulated difference value D the smallest, and this found optimal offset Δt is the "time offset" when the current distortion feature and the voltage abnormal feature appear within the current time window. When the time synchronization verification passes (Δt < preset threshold), the spatial correlation unit is activated. Check whether the specific physical location where mechanical damage is detected (the damage point can be located by analyzing the time difference of vibration signal propagation to different sensors, obtaining a set of three-dimensional coordinates) coincides spatially with the area where current-voltage anomalies are detected (the area where electrical anomalies occur can be located by measuring changes in cable impedance, etc., also obtaining a set of spatial coordinates). The system first obtains the set of position coordinates corresponding to the mechanical damage feature (representing the range of the damage area) and the set of spatial coordinates of the current-voltage abnormal area (representing the range of the electrical abnormal area). Then, it uses the minimum coverage range of the mechanical damage area (such as a cuboid area) as a reference to traverse the coverage ranges of all current-voltage abnormal areas. The system will calculate the overlapping lengths (Lx and Ly) of the mechanical damage area and each electrical abnormal area in the cable length direction (horizontal) and the cable cross-section direction (vertical) respectively.

[0107] Finally, the system needs to comprehensively consider the synchronization degree in time (time offset Δt) and the coincidence degree in space (spatial coincidence degree S) to evaluate the overall correlation strength of the three features of current, voltage, and vibration in this event. A composite feature vector is generated through step 1033. Based on the numerical ranges of the time offset and the spatial coincidence degree, a fuzzy inference system is used to determine the correlation weight coefficients: when the time offset < 3 ms and the spatial coincidence degree > 80%, the weight of the current feature is set to 0.5, the voltage to 0.3, and the vibration to 0.2; when the time offset is 3 - 5 ms and the coincidence degree is 60 - 80%, the weights are adjusted to 0.4, 0.3, and 0.3. Through the weighted summation formula , where α, β, and γ are the weight coefficients, a composite feature vector containing 12-dimensional feature parameters is generated, and each dimensional eigenvalue retains three significant decimal places of precision. That is to say, this correlation strength will be converted into the "weight coefficients" of the three features during the final fusion. The larger the weight, the more important and reliable the feature is in the current event. The determination of the weight coefficients is based on a preset rule (fuzzy inference system): if the time offset is very small (Δt < 3 ms) and the spatial coincidence degree is very high (S > 80%), it is considered that the three are highly correlated. The system uses these weight coefficients to perform a weighted summation operation on the time offset Δt, the spatial coincidence degree S, and the original three feature values. The result of this weighted summation is a numerical combination containing multi-dimensional information, called the "composite feature vector". The finally generated composite feature vector is a 12-dimensional data (containing 12 numerical values), which integrates the original feature values and the verified spatio-temporal correlation information, providing a highly concentrated and reliable feature input for the next step of accurate cable fault diagnosis.

[0108] In practical applications, for example, in an intelligent diagnosis system in a coal mine underground, when an abnormality occurs in the cable of the E12 support section, the system captures the change in the slope of the current wave peak at a sampling frequency of 2000 times per second: 3 positive slope mutations exceeding 80 are detected within a 15-ms time window and 2 negative The negative steep drop, and at the same time, high-frequency harmonics with an energy value 18 times that of the standard working condition appear in the voltage waveform in the 2350 Hz frequency band. The spatio-temporal correlation model aligns the current distortion characteristics, voltage anomaly characteristics containing the timestamp "2023-09-15 14:23:35.782" with the sudden energy increase event of 32 m / s² detected by the vibration sensor at the 850 Hz frequency point on the X-axis. Through a 50 ms sliding time window for phase calibration, it is found that the voltage anomaly characteristics have a 3.2 ms delay relative to the current distortion. After dynamic adjustment, the time offset between the two is reduced to 0.8 ms, which is lower than the preset threshold of 5 ms. The spatial correlation unit then retrieves the cable spatial coordinate data within 3 meters of the E12 support, compares the mechanical damage area with a diameter of 1.2 meters with the area with a diameter of 1.5 meters covered by the current and voltage anomaly signals, measures a horizontal overlap of 0.9 meters and a vertical overlap of 1.1 meters, and calculates a spatial coincidence degree of 83%. The system automatically assigns a time correlation weight of 0.4, a spatial weight of 0.5, and a feature intensity weight of 0.1, generating a composite feature vector containing 12-dimensional parameters such as slope mutation amount, harmonic energy integral, and spectral energy gradient. This vector is visually displayed through the red warning area of the cable three-dimensional model, guiding the inspection personnel to accurately lock the break point of the cable armor layer 1.8 meters east of the E12 support.

[0109] In the overall solution of step 103 above, through the spatio-temporal correlation model, cross-dimensional data fusion is carried out on current distortion characteristics, voltage anomaly characteristics, and mechanical damage characteristics. The dynamic phase alignment algorithm is used to accurately calculate the time offset of current and voltage anomaly characteristics. Combining the overlapping area of the spatial coordinate set for quantitative analysis, a multi-dimensional feature correlation degree weight coefficient is generated. The weighted fusion mechanism integrates time synchronization, spatial coincidence degree, and feature contribution degree into a composite feature vector, effectively solving the misjudgment problem caused by the isolated analysis of electrical and mechanical characteristics in traditional methods. Through the spatio-temporal double verification mechanism, isolated anomaly signals caused by local interference are excluded, enhancing the credibility of fault characteristics under multi-physical field coupling conditions. At the same time, based on the dynamic weight allocation strategy, the importance of features under different working conditions is adaptively adjusted, significantly improving the characterization ability and diagnostic robustness of composite fault characteristics, and providing a strong correlation evidence chain across spatio-temporal dimensions for the accurate identification of early hidden faults of cables.

[0110] 104. Based on the environmental interference factor in the composite feature vector, combined with the cable surface temperature and dust data, dynamically correct the threshold in the fault feature database;

[0111] Optionally, step 104 may specifically include the following steps:

[0112] 1041. Extract the environmental interference factor from the composite feature vector, and the environmental interference factor is calculated by multiplying the temperature value collected by the cable surface temperature sensor by the dust concentration monitoring value;

[0113] 1042. Select a correction factor table corresponding to the temperature compensation coefficient and the dust attenuation coefficient from the fault feature database according to the magnitude of the environmental interference factor;

[0114] 1043. Linearly combine the original threshold in the fault feature database with the temperature compensation coefficient and the dust attenuation coefficient in the correction factor table to generate a dynamically corrected threshold, where the temperature compensation coefficient has a piecewise linear relationship with the temperature value, and the dust attenuation coefficient has an exponential attenuation relationship with the dust concentration value.

[0115] In the above steps, the environmental interference factor refers to an interference intensity index composed of the product of the cable surface temperature and the dust concentration. The temperature compensation coefficient refers to a linear correction parameter for adjusting the feature threshold according to temperature changes. The dust attenuation coefficient refers to an exponential correction parameter reflecting the influence of dust concentration on signal transmission quality. The correction factor table refers to a two-dimensional query table storing the temperature compensation coefficient and the dust attenuation coefficient corresponding to different environmental interference intervals. The piecewise linear relationship means that the temperature compensation coefficient adopts different linear slopes in different temperature intervals. The exponential attenuation relationship means that the dust attenuation coefficient decreases according to the mathematical relationship of the function law with the increase of dust concentration.

[0116] In the embodiments of the present application, in order to make the judgment criteria (thresholds) for fault diagnosis more intelligent and adaptable to the actual environment, and to avoid misjudgment or missed judgment caused by environmental interference (mainly temperature and dust). The environmental conditions where the cable is located (such as the surface temperature and the dust concentration) will directly affect the signal quality collected by the sensor and the manifestation intensity of the fault characteristics. Therefore, the system needs to dynamically adjust the "standard lines" (thresholds) stored in the fault characteristic database for judging whether a fault occurs according to the real-time degree of environmental interference. First, calculate the environmental interference factor through step 1041. The cable surface temperature sensor uses an infrared temperature measurement module to collect temperature data at a frequency of 10 times per second, and the dust concentration monitor outputs a minute-level average value based on the principle of light scattering. The temperature value (unit: °C) and the dust concentration value (unit: mg / m³) are normalized to eliminate the dimension difference, and then the environmental interference factor is calculated through the product formula EIF = 0.7T + 0.3D, where T is the normalized temperature value and D is the normalized dust concentration value. The coefficients 0.7 and 0.3 are empirical weight values, indicating that the influence weight of temperature on interference (70%) is greater than that of dust (30%). That is to say, the system needs to quantify the current degree of environmental interference, which is achieved through an index called "environmental interference factor". This factor is calculated from two key environmental parameters: one is the real-time temperature on the cable surface (collected by an infrared temperature sensor, such as measuring 10 times per second), and the other is the dust concentration in the environment (monitored by a dust sensor based on the principle of light scattering, usually outputting an average value per minute). In order to fairly combine these two quantities with different units (temperature is in degrees Celsius °C, and dust is in milligrams per cubic meter mg / m³), the system will first normalize them (imagine scaling them proportionally to between 0 and 1 or a certain common range). For example, if the measured cable surface temperature is 60 °C, after normalization, T = 0.6 (assuming the normalization range is 0 - 100 °C); the dust concentration is 150 mg / m³, and after normalization, D = 0.3 (assuming the normalization range is 0 - 500 mg / m³). Then the environmental interference factor EIF = 0.7 * 0.6 + 0.3 * 0.3 = 0.42 + 0.09 = 0.51.

[0117] Secondly, after having the environmental interference factor EIF, the system knows the "level" of the current environmental interference. Next, the system will look up the corresponding "correction coefficient table" in the fault characteristic database according to the size of this EIF value. Select the correction coefficient table through step 1042. The fault characteristic database is divided into five intervals according to the numerical range of the environmental interference factor: when EIF < 0.3, select the low-temperature and low-dust correction table; when 0.3 ≤ EIF < 0.6, select the normal-temperature and normal-convention table; when 0.6 ≤ EIF < 0.9, select the high-temperature and low-dust table; when 0.9 ≤ EIF < 1.2, select the high-dust compensation table; when EIF ≥ 1.2, select the extreme environment table. Each correction coefficient table contains a temperature compensation coefficient and a dust attenuation coefficient The two-dimensional matrix, where the rows of the matrix correspond to the segmented nodes of the temperature compensation coefficient every 5°C in the range of 20°C to 80°C, and the columns correspond to the segmented nodes of the dust concentration every 50 mg / m³ in the range of 0 - 500 mg / m³. That is to say, the database has pre-divided five intervals according to different interference levels, and prepared a special correction coefficient table for each interval: when the EIF is very low (<0.3), select the "low temperature and low dust correction table"; when the EIF is relatively low (0.3 <= EIF < 0.6), select the "normal temperature and normal condition table"; when the EIF is medium (0.6 <= EIF < 0.9), select the "high temperature and low dust table"; when the EIF is relatively high (0.9 <= EIF < 1.2), select the "high dust compensation table"; when the EIF is very high (>=1.2), then select the "extreme environment table". These correction coefficient tables store two very important correction coefficients: one is the "temperature compensation coefficient" α(T), and the other is the "dust attenuation coefficient" β(D). Each table is a two-dimensional matrix, with rows corresponding to different temperature segmentation points (such as from 20°C to 80°C, one node every 5°C), and columns corresponding to different dust concentration segmentation points (such as from 0 mg / m³ to 500 mg / m³, one node every 50 mg / m³). The system can find the closest temperature node and dust concentration node in the selected correction coefficient table according to the currently measured temperature and dust concentration values, and then look up the corresponding α(T) and β(D) values. For example, when the current EIF = 0.51, which belongs to the interval [0.3, 0.6), select the "normal temperature and normal condition table". The measured temperature is 45°C, and look up the row corresponding to the temperature node 40°C or 50°C in the table (interpolation may be required); the measured dust concentration is 120 mg / m³, and look up the column corresponding to the concentration node 100 mg / m³ or 150 mg / m³ in the table (interpolation may be required), and finally obtain the specific values of α(T) and β(D).

[0118] Finally, it is to use the temperature compensation coefficient α(T) and dust attenuation coefficient β(D) obtained by looking up the table to dynamically correct the original threshold θ0 in the fault feature database to obtain a new threshold suitable for the current environment 。Dynamically correct the threshold through step 1043. For the original fault threshold Perform double coefficient correction calculation: 。The temperature compensation coefficient Linearly increases at a rate of 0.02 / °C in the range of 20 - 50°C and at a rate of 0.05 / °C in the range of 50 - 80°C. The dust attenuation coefficient Adopts The exponential function calculation of, and the attenuation coefficient reaches 0.632 when D = 200 mg / m³. The corrected threshold Both the positive offset brought about by temperature compensation and the negative compensation caused by dust attenuation are retained, and finally a dynamic threshold parameter group adapted to the actual environmental conditions is generated. That is to say, an increase in temperature usually makes some fault characteristics (such as resistance, certain signal amplitudes) more obvious. Therefore, a positive compensation (+α(T)) is required to increase the threshold (make the diagnostic criterion slightly looser to avoid misjudging the increase in temperature itself as a fault). The growth rate (slope) of the temperature compensation coefficient α(T) is different in different temperature ranges: in a mild temperature range (such as 20°C to 50°C), for every 1°C increase in temperature, α(T) increases by a relatively small proportion (such as 0.02); in a high temperature range (such as 50°C to 80°C), for every 1°C increase in temperature, α(T) increases by a relatively large proportion (such as 0.05), which reflects that the influence of temperature on the system may not be linear, and the influence is greater at high temperatures. For example, in the temperature range below 50°C, α(T)=0.02*(T - 20) (assuming α = 0 at 20°C). If T = 40°C, then α(T)=0.02*(40 - 20)=0.4. On the other hand, an increase in dust concentration usually interferes with signal transmission and weakens the intensity of the fault characteristics detected by the sensor. Therefore, a negative attenuation (-β(D)) is required to reduce the threshold (make the diagnostic criterion slightly stricter to avoid missing a judgment due to the weakening of the true fault signal caused by dust occlusion). The dust attenuation coefficient β(D) increases with the increase in dust concentration D, and it has an exponential growth relationship: β(D)=1 - e^(-0.005*D). The characteristic of the exponential function is that when the dust concentration is very low, the attenuation is very small; the higher the concentration, the faster the attenuation increases, approaching 1 (i.e., the maximum attenuation of 100%). For example, when D = 0mg / m³, β(D)=1 - e^0=1 - 1=0; when D = 100mg / m³, β(D)=1 - e^(-0.005*100)=1 - e^(-0.5)≈1 - 0.6065=0.3935; when D = 200mg / m³, β(D)=1 - e^(-1)≈1 - 0.3679=0.6321. Finally, the original threshold θ0 is amplified by (1 + α(T)) (temperature compensation) and reduced by (1 - β(D)) (dust attenuation), and a dynamic new threshold adapted to the current specific environmental conditions (temperature T and dust concentration D) is generated .

[0119] In practical applications, for example, in an intelligent diagnosis system in a certain coal mine underground, when the surface temperature sensor of the cable in section E12 detects a high temperature of 65°C and the dust concentration monitor shows 220 mg / m³, the system automatically calculates the environmental interference factor as 14300 (65×220). The fault feature database calls a preset correction coefficient table: when the interference factor exceeds 12000, the second correction scheme is enabled. The temperature compensation coefficient is taken as 0.8, corresponding to the temperature range of 60 - 70°C, and the dust attenuation coefficient is taken as 0.7, following the exponential decay curve for the concentration range of 200 - 250 mg / m³. The system corrects the current mutation threshold of 85A and the voltage harmonic threshold of 50 mV in the original fault threshold database to 85×0.8×0.7 = 47.6A and 50×0.8×0.7 = 28 mV respectively, and at the same time adjusts the vibration energy threshold from 25 m / s² to 25×0.7 = 17.5 m / s². This dynamic threshold mechanism enables the system to accurately distinguish between normal load fluctuations and real short - circuit faults when the shearer cutting generates an 80A current fluctuation, avoiding false alarms caused by changes in cable impedance due to high temperature.

[0120] In the overall scheme of step 104 above, by fusing the monitoring data of the cable surface temperature and dust concentration to construct the environmental interference factor, combining the piece - wise linear adjustment mechanism of the temperature compensation coefficient and the exponential decay model of the dust attenuation coefficient to dynamically correct the fault feature thresholds, using the product calculation method to quantify the environmental interference intensity and adaptively match the correction coefficient table, and using the linear combination algorithm to dynamically couple the environmental parameters with the original thresholds, it effectively solves the problem of misjudgment of features caused by temperature drift and dust shielding of traditional fixed thresholds under complex working conditions. Through the compensation mechanism, it improves the adaptability of the fault diagnosis system to changes in the cable surface state, significantly reduces the false alarm rate and missed detection probability in high - temperature and high - dust environments, and at the same time enhances the robustness and self - healing ability of the feature database under different environmental interference intensities, ensuring the reliability of the cable fault diagnosis results and the accuracy of maintenance decisions.

[0121] 105. Perform multi - dimensional matching calculations on the composite feature vector and the corrected fault mode thresholds, and determine the fault diagnosis result according to the correlation strength between the electrical features and mechanical features in the matching result.

[0122] Optionally, step 105 may specifically include the following steps:

[0123] 1051. Decompose the composite feature vector into an electrical feature sub - vector and a mechanical feature sub - vector, and perform Euclidean distance calculations on them respectively with the corrected fault mode thresholds to obtain the electrical matching degree and the mechanical matching degree;

[0124] 1052. Calculate the correlation strength value between the electrical matching degree and the mechanical matching degree. When the correlation strength value exceeds a preset fusion threshold, activate the fault type determination module to compare the deviation between the electrical matching degree and the short - circuit spike threshold with the deviation between the mechanical matching degree and the open - circuit oscillation threshold through the fault type determination module, and determine the fault diagnosis result according to the comparison result.

[0125] Among them, step 1052 may specifically include the following process: If the deviation between the electrical matching degree and the short - circuit spike threshold is less than the deviation between the mechanical matching degree and the open - circuit oscillation threshold, output the short - circuit spike diagnosis result corresponding to cable damage; if the deviation between the electrical matching degree and the short - circuit spike threshold is greater than the deviation between the mechanical matching degree and the open - circuit oscillation threshold, output the open - circuit oscillation diagnosis result corresponding to electrical abnormality.

[0126] In the above steps, the composite feature vector refers to a multi - dimensional data set containing current distortion, voltage abnormality, and mechanical damage features. The electrical matching degree refers to the similarity measure between the electrical feature sub - vector in the composite feature vector and the fault threshold. The mechanical matching degree refers to the similarity measure between the mechanical feature sub - vector and the fault threshold. The correlation strength value refers to the product or weighted calculation result of the electrical matching degree and the mechanical matching degree. The fusion threshold refers to the critical correlation strength value for triggering fault type determination. The fault type determination module refers to a decision - making unit that determines the fault type based on the comparison of electrical and mechanical feature deviations. The short - circuit spike threshold refers to the current - voltage feature matching threshold corresponding to cable short - circuit faults. The open - circuit oscillation threshold refers to the mechanical vibration feature matching threshold corresponding to line open - circuit faults.

[0127] In the embodiments of the present application, the "composite feature vector" that integrates spatio-temporal correlation information generated in the previous steps is comprehensively compared with the "fault mode threshold" dynamically adjusted according to environmental conditions, so as to determine the specific fault type (such as short circuit or open circuit) of the cable and its location. The composite feature vector is decomposed into a part representing electrical anomalies (electrical feature sub-vector) and a part representing mechanical damage (mechanical feature sub-vector), and similarity matching calculations are respectively performed with the corresponding fault criteria (modified thresholds) to obtain the "electrical matching degree" and the "mechanical matching degree". Then, the system will pay special attention to the correlation strength between these two matching degrees (that is, whether they are both very high or one is very high and the other is very low). Only when the correlation strength is large enough can it be considered that the fault type can be reliably determined. First, feature matching calculations are performed through step 1051. The composite feature vector is decomposed into an electrical feature sub-vector containing the change amount of the wave peak slope and the energy in the harmonic frequency band, and a mechanical feature sub-vector containing the mechanical damage spectrum features according to the feature type. The Euclidean distance formula is used to calculate the geometric distance between the electrical feature sub-vector and the modified short-circuit spike threshold, and the calculation result is normalized to obtain the electrical matching degree in the range of 0-1. The geometric distance between the mechanical feature sub-vector and the modified open-circuit oscillation threshold is synchronously calculated, and the mechanical matching degree is obtained through the same normalization process. That is to say, the system splits the composite feature vector (F) containing multiple values into two parts according to the nature of the features: one part is the "electrical feature sub-vector" (E), which mainly contains the two indicators reflecting electrical anomalies, namely the change amount of the current wave peak slope and the voltage harmonic frequency band energy calculated previously; the other part is the "mechanical feature sub-vector" (M), which mainly contains the frequency points with sudden energy increase and their amplitude increase information in the mechanical damage spectrum feature set generated previously, reflecting physical structure damage. Then, the system compares the electrical feature sub-vector (E) with the standard threshold vector (such as containing the current slope threshold, voltage harmonic energy threshold, etc.) predefined for the "short-circuit spike fault" in the fault feature database and corrected by the environment, and calculates the "Euclidean distance" between them (a mathematical method to measure the distance between two points in a multi-dimensional space, the smaller the distance, the more similar). For easier understanding, this distance is converted into an "electrical matching degree" between 0% and 100% (the higher the percentage, the more the current electrical feature resembles a typical short-circuit fault). Similarly, the system compares the mechanical feature sub-vector (M) with the standard threshold vector (such as containing the vibration energy threshold at a specific frequency point, etc.) predefined for the "open-circuit oscillation fault" and corrected by the environment, calculates the Euclidean distance and converts it into a "mechanical matching degree" between 0% and 100% (the higher the percentage, the more the current mechanical vibration feature resembles a typical open-circuit fault). For example, the composite feature vector F is decomposed, the electrical sub-vector E contains the current slope change amount = 82 A / ms, the voltage harmonic energy = 480 mV² / Hz; the mechanical sub-vector M contains the vibration energy at the 850 Hz frequency point = 28 m / s².The corrected short - circuit spike threshold is: current slope 70 A / ms, harmonic energy 350 mV² / Hz. The Euclidean distance between E and the short - circuit threshold is calculated to be 15.2, and it is assumed that the electrical matching degree after normalization is 78%. The corrected open - circuit oscillation threshold is: 850 Hz vibration energy = 20 m / s². The Euclidean distance between M and the open - circuit threshold is calculated to be 8, and it is assumed that the mechanical matching degree after normalization is 60%.

[0128] Secondly, after obtaining the electrical matching degree and the mechanical matching degree, the system needs to determine whether these two results "cooperatively" point to the same fault event, which is achieved by calculating their "correlation strength value". When the correlation strength is high enough to trigger the fault type determination module, the fault type determination module is a decision-making unit that determines the fault type based on the comparison of electrical and mechanical characteristic deviations. This module will compare the "deviation" between the electrical matching degree and the short-circuit fault standard (short-circuit peak threshold) and the "deviation" between the mechanical matching degree and the open-circuit fault standard (open-circuit oscillation threshold). The "deviation" here refers to the original Euclidean distance value (or its absolute value) obtained when calculating the matching degree earlier. The correlation strength determination is performed through step 1052. The electrical matching degree and the mechanical matching degree are weighted and multiplied to calculate the correlation strength value, and the weight coefficient is dynamically adjusted according to the environmental interference factor. When the correlation strength value exceeds the preset fusion threshold of 0.75, the fault type determination module is activated. This module calculates the absolute deviation value ΔE between the electrical matching degree and the short-circuit peak threshold, and the absolute deviation value ΔM between the mechanical matching degree and the open-circuit oscillation threshold respectively. The final fault type is determined by comparing the magnitudes of ΔE and ΔM: if ΔE is less than ΔM, it is determined to be a short-circuit peak fault caused by cable insulation damage; if ΔE is greater than ΔM, it is determined to be an open-circuit oscillation fault caused by mechanical fracture, and a diagnostic result report including the fault location coordinates and the confidence rating is generated. That is to say, the system presets a "fusion threshold" (such as 75%). Only when the calculated correlation strength value exceeds this threshold (for example, in a certain case, the correlation strength reaches 85% > 75%), the system believes that there is a strong enough correlation between the electrical anomaly and the mechanical damage (most likely caused by the same fault source), and activates the final "fault type determination module" for precise classification. If the correlation strength value is lower than the threshold, it indicates that the electrical and mechanical characteristics may be out of sync or irrelevant, and the system may need to collect more data or give an uncertain prompt. Compare the "deviation value (ΔE) between the electrical matching degree and the short-circuit peak threshold" and the "deviation value (ΔM) between the mechanical matching degree and the open-circuit oscillation threshold". If the deviation of the electrical part is smaller (ΔE < ΔM), it means that the current electrical characteristics are more in line with the typical mode of short-circuit faults (even if the mechanical characteristics also have a certain degree of matching, but not as typical as the electrical characteristics), and the system determines it as a "short-circuit peak fault caused by cable insulation damage". On the contrary, if the deviation of the mechanical part is smaller (ΔE > ΔM), it means that the current mechanical vibration characteristics are more in line with the typical mode of open-circuit faults, and the system determines it as an "open-circuit oscillation fault caused by mechanical fracture or looseness". Continuing with the above example: the electrical matching degree is 78% corresponding to the original Euclidean distance ΔE = 15.2 (the distance from the short-circuit threshold), and the mechanical matching degree is 60% corresponding to the original Euclidean distance ΔM = 8 (the distance from the open-circuit threshold). Comparing ΔE = 15.2 and ΔM = 8, it is found that ΔE > ΔM (15.2 > 8), which means that the deviation of the mechanical characteristics relative to its standard (open-circuit threshold) is smaller and more typical.Therefore, although a high correlation strength indicates the existence of a problem, according to the deviation comparison rule, the system finally determines it as "the open - circuit oscillation diagnosis result corresponding to electrical anomalies" (Note: This determination result needs to be combined with specific threshold definitions and scenarios, and is an example according to the original text rules here). Finally, the system will generate a comprehensive report containing the determined fault type (such as "short - circuit spike" or "open - circuit oscillation"), the precise location of the fault occurrence (three - dimensional coordinates), and the credibility of the diagnosis result (confidence level).

[0129] In practical applications, for example, in an intelligent cable diagnosis system in a coal mine underground, when the system detects an anomaly in the cable of the E12 support section, the collected composite feature vector includes the change in the current wave - peak slope of 82 A / ms, the harmonic energy value of 480 mV² / Hz in the 2350 Hz frequency band of the voltage, and the energy value of 28 m / s² at the 850 Hz frequency point in the vibration spectrum. The system decomposes this vector into an electrical sub - vector (including the first two items) and a mechanical sub - vector (the third item), and respectively performs matching calculations with the dynamically corrected fault thresholds: The Euclidean distance between the electrical sub - vector and the short - circuit spike thresholds (current slope threshold of 70 A / ms, harmonic energy threshold of 350 mV² / Hz) is 15.2, and the matching degree reaches 78%; the Euclidean distance between the mechanical sub - vector and the open - circuit oscillation threshold (vibration energy threshold of 20 m / s²) is 8, and the matching degree is 60%. When the system calculates the correlation strength value of the two, it is found that when there is a 0.5 - ms time synchronization between the vibration signal generated by the shearer pick hitting the gangue and the current mutation, the correlation strength value reaches 0.85, exceeding the preset fusion threshold of 0.75. At this time, the fault determination module starts to compare: The deviation between the electrical matching degree and the short - circuit threshold is 15.2 - 12.8 = 2.4, and the deviation between the mechanical matching degree and the open - circuit threshold is 8 - 5 = 3. Since the electrical deviation is smaller, the system determines it as "discharge short - circuit caused by cable insulation damage", and accurately locates a 3 - cm - long crack in the cable outer skin 1.8 meters east of the E12 support in the three - dimensional model. This diagnosis result enables the maintenance personnel to directly reach the target area with a partial discharge detector and confirm the fault point within 10 minutes, saving 4 hours of operation time compared with the traditional full - line inspection.

[0130] In the overall solution of step 105 above, accurate discrimination of cable faults is achieved through multi-dimensional decomposition of the composite feature vector and matching calculation of the dynamically corrected threshold. The independent matching degree calculation of the electrical feature sub-vector and the mechanical feature sub-vector and the fusion analysis of the correlation strength value are adopted. Combining with the deviation comparison mechanism of the fault type determination module, different diagnostic conclusions are output based on the deviation magnitude difference between the short-circuit spike threshold and the open-circuit oscillation threshold, effectively solving the problem of fault type confusion caused by feature coupling in traditional methods. The risk of false triggering of a single feature is excluded through the collaborative verification of electrical and mechanical features, significantly enhancing the ability to distinguish short-circuit spikes from mechanical damage oscillations. At the same time, the multi-dimensional matching based on the dynamically corrected threshold enhances the adaptability and decision credibility of the diagnostic system for composite faults, providing an intelligent diagnostic solution with both sensitivity and specificity for mine cable operation and maintenance.

[0131] The following is a complete embodiment for steps 101 to 105:

[0132] As Figure 2As shown, in a coal mine underground power supply system, intelligent monitoring devices arranged along the coal mining face collect cable operation data. When the coal shearer cuts gangue, the system first synchronously collects cable operation data through a multi-dimensional sensor group (step 101): the high-frequency current sensor detects that the current in phase C suddenly increases from 520 A to 720 A within 0.5 seconds, and at the same time the voltage drops suddenly from 380 V to 260 V, and the vibration sensor captures an instantaneous impact of 3.2 g in the vertical direction. Facing the humid and dusty underground environment (dust concentration up to 280 milligrams per cubic meter - mg / m³), the system starts adaptive noise reduction and time sequence synchronous transmission. The system filters out the 12 kHz high-frequency noise generated by the frequency converter through a Butterworth filter with an 8 kHz cut-off frequency, and starts the redundant transmission mode when the dust concentration reaches 280 mg / m³, compressing the signal interval to 15 ms to ensure that the time stamp deviation between the vibration and the electrical signal is less than ±0.8 ms. These signals that have been noise-reduced and strictly time-aligned are then superimposed and analyzed to generate a high-quality anti-interference data set. Based on this data set, the system enters the feature extraction stage. It is found that there is a significant correlation between the violent vibration of the cable (high intensity) and the abnormally rapid change of the current (transient rate up to 150 amperes per millisecond - A / ms). The dynamic capture window is extended to 40 ms according to the correlation between the vibration intensity and the current transient rate (150 A / ms), and three peak transitions are captured. Subsequently, it accurately calculates that the slope change of the largest one of these peaks reaches 86.7 A / ms (describing the steepness of the current climbing from one peak to the next). At the same time, the frequency domain analysis is performed on the voltage waveform segment within the same time window, the maximum slope change of 86.7 A / ms is extracted, and a harmonic energy peak of 450 mV² / Hz is detected in the 2350 Hz frequency band of the voltage waveform. At the same time, wavelet packet decomposition of the vibration signal finds that the energy at the 850 Hz frequency point soars from 5 m / s² to 25 m / s², forming a mechanical damage spectrum feature.

[0133] Next, the system inputs these key features (current peak slope of 86.7 A / ms, voltage 2350 Hz harmonic energy of 450 mV² / Hz, vibration 850 Hz energy peak of 25 m / s²) into the spatio-temporal correlation model. The spatio-temporal correlation model verifies that there is a 1.2 ms time offset between the current distortion and the voltage anomaly and an 85% spatial coincidence within a 2 m range near the E12 support. Combining the cable surface temperature of 68 °C and the dust concentration of 280 mg / m³, the environmental interference factor is calculated as 19040, and the short-circuit current threshold is dynamically corrected to 52 A and the harmonic energy threshold to 210 mV² / Hz. The model first verifies the temporal synchronization of the electrical anomaly features: it confirms that the time points at which the current distortion feature and the voltage anomaly feature appear differ by only 1.2 milliseconds (ms), much less than the preset tolerance threshold of 5 ms, indicating a high correlation between the two. Immediately afterwards, the spatial position comparison shows that the position of the detected mechanical damage feature (located by the vibration source) and the detected electrical anomaly area (located by the electrical signal feature) coincide spatially by up to 85% within a 2 m range near the E12 hydraulic support. Based on this high spatio-temporal correlation, the model generates a composite feature vector that combines electrical and mechanical anomaly information. At the same time, the system monitors the environmental conditions in real time. The cable surface temperature sensor measures 68 degrees Celsius (°C), and the dust concentration monitoring value remains at a high level of 280 mg / m³. The environmental interference factor (EIF) calculated by combining these two environmental parameters is as high as 19040 (quantifying the degree of interference in the harsh environment). Based on this high interference factor, the system dynamically adjusts the threshold criteria in the fault feature database: the current mutation threshold for judging a short-circuit fault is corrected and reduced to 52 amperes (A), and the harmonic energy threshold is corrected and reduced to 210 millivolt square per hertz (mV² / Hz). This correction takes into account the actual situation that high temperature and dust will amplify certain electrical signal features and attenuate sensor signals respectively, making the diagnostic criteria more adaptable to the harsh underground environment. Finally, the system performs multi-dimensional matching calculations on the generated composite feature vector and these dynamically corrected thresholds. The calculation results show that the matching degree of the electrical features (current slope, voltage harmonics) with the corrected short-circuit fault threshold reaches 82%, and the matching degree of the mechanical feature (850 Hz vibration energy) with the corrected open-circuit fault threshold reaches 73%. More importantly, the correlation strength value (comprehensively reflecting the likelihood of the co-occurrence of electrical and mechanical anomalies) is as high as 0.89, significantly exceeding the preset fusion trigger threshold of 0.75. According to the judgment rule, the system further compares the deviation of the electrical feature matching degree from the short-circuit threshold (smaller) and the deviation of the mechanical feature matching degree from the open-circuit threshold (relatively larger), and finally determines the fault type as: an intermittent short-circuit fault caused by mechanical damage (such as extrusion, scratching) to the cable insulation layer. The system combines spatio-temporal positioning information and accurately locates the fault point in the cable section 2.3 m east of the E12 hydraulic support in the three-dimensional model.On-site maintenance personnel went directly to the target location based on this diagnostic result and indeed found a crack about 5 centimeters (cm) long in the cable insulation layer. They verified the accuracy of the diagnosis through insulation resistance testing. Multidimensional matching showed an electrical feature matching degree of 82% and a mechanical matching degree of 73%. The correlation strength value of 0.89 exceeded the threshold, and it was determined that the intermittent short circuit was caused by mechanical damage to the cable insulation layer. The three-dimensional model located the fault point 2.3 meters east of the E12 bracket. On-site maintenance found a 5-cm-long insulation crack, and the diagnosis was verified accurate through insulation testing. The overall troubleshooting time was significantly shortened, effectively avoiding unplanned shutdown accidents. This precise positioning significantly shortened the overall fault troubleshooting time, effectively avoiding unplanned shutdown accidents caused by cable failures and ensuring continuous safe production underground in coal mines.

[0134] Figure 3 The following is a schematic structural diagram of a mine circuit fault self-diagnosis system provided by an embodiment of the present application. As Figure 2 shown, the system includes:

[0135] An acquisition module 31 that synchronously acquires the current waveform, voltage waveform, and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generates an anti-interference data set through adaptive noise reduction and time-sequence synchronization transmission in a humid and dusty environment;

[0136] An extraction module 32 that dynamically adjusts the waveform interception window to capture transient waveform segments based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference data set, extracts the change amount of the wave peak slope from the transient waveform segments of the current waveform, separates the harmonic frequency band energy of the short-circuit spike from the transient waveform segments of the voltage waveform, and simultaneously performs time-frequency analysis on the three-dimensional vibration signal to generate a mechanical damage spectrum feature set;

[0137] A correlation module 33 that inputs the change amount of the wave peak slope, the harmonic frequency band energy, and the mechanical damage spectrum feature set into a spatio-temporal correlation model, and generates a composite feature vector through spatio-temporal correlation verification of the current distortion feature and the voltage anomaly feature;

[0138] A correction module 34 that dynamically corrects the threshold in the fault feature database based on the environmental interference factor in the composite feature vector, in combination with the cable surface temperature and dust data;

[0139] A matching module 35 that performs multi-dimensional matching calculations on the composite feature vector and the corrected fault mode threshold, and determines the fault diagnosis result according to the correlation strength between the electrical feature and the mechanical feature in the matching result.

[0140] Figure 3 The mine circuit fault self-diagnosis system described above can execute Figure 1The implementation principle and technical effects of the mine circuit fault self-diagnosis method described in the illustrated embodiments will not be elaborated further. For the mine circuit fault self-diagnosis system in the above embodiments, the specific ways in which each module and unit perform operations have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A self-diagnosis method for mine circuit faults, characterized in that, Including: Synchronously collect the current waveform, voltage waveform and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generate an anti-interference data set through adaptive noise reduction and time-sequence synchronous transmission in a humid and dusty environment; Based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference data set, dynamically adjust the waveform truncation window to capture transient waveform segments, extract the change amount of the wave peak slope from the transient waveform segments of the current waveform, separate the harmonic band energy of the short-circuit spike from the transient waveform segments of the voltage waveform, and at the same time perform time-frequency analysis on the three-dimensional vibration signal to generate a mechanical damage spectrum feature set; Input the change amount of the wave peak slope, the harmonic band energy and the mechanical damage spectrum feature set into the spatio-temporal correlation model, and generate a composite feature vector through the spatio-temporal correlation verification of the current distortion feature and the voltage anomaly feature; Based on the environmental interference factor in the composite feature vector, combined with the cable surface temperature and dust data, dynamically correct the threshold in the fault feature database; Perform multi-dimensional matching calculations on the composite feature vector and the corrected fault mode threshold, and determine the fault diagnosis result according to the correlation strength between the electrical feature and the mechanical feature in the matching result.

2. The method according to claim 1, wherein Perform multi-dimensional matching calculations on the composite feature vector and the corrected fault mode threshold, and determine the fault diagnosis result according to the correlation strength between the electrical feature and the mechanical feature in the matching result, including: Decompose the composite feature vector into an electrical feature sub-vector and a mechanical feature sub-vector, and perform Euclidean distance calculations on them respectively with the corrected fault mode threshold to obtain the electrical matching degree and the mechanical matching degree; Calculate the correlation strength value between the electrical matching degree and the mechanical matching degree. When the correlation strength value exceeds the preset fusion threshold, activate the fault type determination module to compare the deviation between the electrical matching degree and the short-circuit spike threshold and the deviation between the mechanical matching degree and the open-circuit oscillation threshold through the fault type determination module, and determine the fault diagnosis result according to the comparison result.

3. The method according to claim 2, wherein The determining the fault diagnosis result according to the comparison result includes: If the deviation between the electrical matching degree and the short-circuit spike threshold is less than the deviation between the mechanical matching degree and the open-circuit oscillation threshold, output the short-circuit spike diagnosis result corresponding to the cable damage; If the deviation between the electrical matching degree and the short-circuit spike threshold is greater than the deviation between the mechanical matching degree and the open-circuit oscillation threshold, output the open-circuit oscillation diagnosis result corresponding to the electrical anomaly.

4. The method according to claim 1, wherein Based on the environmental interference factor in the composite feature vector, combined with the cable surface temperature and dust data, dynamically correct the threshold in the fault feature database, including: Extract the environmental interference factor from the composite feature vector, and the environmental interference factor is calculated by the product of the temperature value collected by the cable surface temperature sensor and the dust concentration monitoring value; According to the size of the environmental interference factor, select the correction coefficient table containing the temperature compensation coefficient and the dust attenuation coefficient from the fault feature database; The original threshold in the fault feature database is linearly combined with the temperature compensation coefficient and the dust attenuation coefficient in the correction coefficient table to generate a dynamically corrected threshold, wherein the temperature compensation coefficient is in a piecewise linear relationship with the temperature value, and the dust attenuation coefficient is in an exponential attenuation relationship with the dust concentration value.

5. The method according to claim 1, characterized in that The peak slope variation, harmonic frequency band energy and mechanical damage spectrum feature set are input into the spatiotemporal correlation model, and a composite feature vector is generated through spatiotemporal correlation verification of current distortion features and voltage anomaly features, including: The peak slope variation, harmonic frequency band energy and mechanical damage spectrum feature set are input into a feature fusion module, and data is reorganized according to a timestamp alignment rule to form a spatiotemporal matrix including current distortion features, voltage anomaly features and mechanical damage features; Based on the current distortion characteristics and voltage anomaly characteristics in the space-time matrix, the time offset of the two within a preset time window is calculated by a cross-validation unit in the space-time association model. When the time offset is less than a preset threshold, the space association unit is triggered to compare the physical position corresponding to the mechanical damage characteristics with the spatial coordinates of the current and voltage anomaly area to generate a spatial coincidence; According to the time offset and spatial overlap, the correlation weight coefficients of the current distortion feature, the voltage abnormality feature and the mechanical damage feature are determined, and the time offset, the spatial overlap and the correlation weight coefficients are weighted and summed to generate a composite feature vector.

6. The method according to claim 5, wherein Based on the current distortion characteristics and voltage anomaly characteristics in the space-time matrix, the time offset of the two in a preset time window is calculated by a cross-validation unit in the space-time association model. When the time offset is less than a preset threshold, the space association unit is triggered to compare the physical position corresponding to the mechanical damage characteristics with the spatial coordinates of the current and voltage anomaly area to generate a spatial coincidence, including: Extracting a current distortion feature sequence and a voltage anomaly feature sequence from the space-time matrix, segmenting and intercepting the current distortion feature sequence and the voltage anomaly feature sequence according to a preset time window sliding rule to form a current distortion feature segment and a voltage anomaly feature segment containing feature values of a plurality of consecutive sampling points; Performing point-by-point difference accumulation calculation on the current distortion characteristic segment and the voltage anomaly characteristic segment in the same time window, and dynamically adjusting the starting position of the voltage anomaly characteristic segment to minimize the phase difference between the two, thereby generating a time offset in the current time window; Determine whether the time offset is less than a preset threshold value, and when the time offset is less than the preset threshold value, send a trigger instruction to the space association unit, and retrieve the physical position coordinate set corresponding to the mechanical damage feature and the space coordinate set of the current and voltage abnormal area through the trigger instruction; Based on the physical position coordinate set and the spatial coordinate set of the current and voltage abnormal areas, taking the minimum coverage area of the mechanical damage feature as a benchmark, traverse the coverage areas of all current and voltage abnormal areas, and count the overlapping lengths of the two in the horizontal and vertical ranges. Generate a spatial overlap based on the ratio of the product of the horizontal and vertical overlapping lengths to the coverage area of the mechanical damage feature.

7. The method according to claim 1, wherein Based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference dataset, dynamically adjust the waveform intercept window to capture transient waveform segments, including: Based on the vibration intensity detection module of the three-dimensional vibration signal in the anti-interference dataset, calculate the root mean square value of the vibration amplitude, and establish a correlation relationship table between the vibration intensity and the current transient rate; When it is detected that the current transient rate exceeds the rate threshold corresponding to the current vibration intensity interval in the correlation relationship table, select the slope mutation point or local extreme point within a preset time range from the current current waveform as the starting position of the waveform intercept window, where the wave valley point is preferentially selected as the starting position; Dynamically adjust the length of the waveform intercept window according to the ratio of the root mean square value to the current transient rate, where the window length is extended when the ratio increases and contracted when the ratio decreases, and the window length is limited between a preset upper limit and a lower limit; Analyze the amplitude mutation amplitude and spectral energy distribution of the current waveform within the adjusted window. If the value of the amplitude mutation amplitude exceeds the preset threshold and the spectral energy is concentrated in the high-frequency band, capture the current waveform as a transient waveform segment.

8. The method according to claim 1, characterized in that, Extract the peak slope change amount from the transient waveform segment of the current waveform, separate the harmonic frequency band energy of the short-circuit spike from the transient waveform segment of the voltage waveform, and simultaneously perform time-frequency analysis on the three-dimensional vibration signal to generate a mechanical damage spectrum feature set, including: Extract the ratio of the amplitude increment to the time increment between adjacent peaks from the captured transient waveform segment of the current waveform as the peak slope change amount; Perform frequency-domain decomposition on the captured transient waveform segment of the voltage waveform, identify the frequency band with a frequency higher than three times the fundamental frequency and a concentrated energy distribution, and calculate the energy integral of each sub-band within the selected frequency band as the harmonic frequency band energy; Perform time-frequency analysis on the three-dimensional vibration signal using a multi-resolution decomposition method, extract the set of frequency points with a sudden increase in vibration energy within each decomposed frequency band, record the frequency position and energy increase amplitude of the set of frequency points, and generate a mechanical damage spectrum feature set composed of frequency point distribution and energy change.

9. The method according to claim 1, characterized in that, Synchronously collect the current waveform, voltage waveform, and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generate an anti-interference dataset through adaptive noise reduction and time-sequence synchronous transmission in a humid and dusty environment, including: Collect the current waveform, voltage waveform, and three-dimensional vibration signal of the cable in a synchronous trigger manner through a sensor group composed of a current sensor, a voltage sensor, and a three-dimensional vibration sensor; According to the signal-to-noise ratio of the current waveform and the voltage waveform, dynamically select filtering parameters through the signal intensity adaptive module, and perform high-frequency noise suppression on the current waveform and the voltage waveform based on the filtering parameters, only retaining the transient components with an amplitude fluctuation exceeding the preset threshold in the signal; When transmitting the current waveform, voltage waveform and three-dimensional vibration signal, the time stamps of the three axial components of the three-dimensional vibration signal, the current waveform and the voltage waveform are aligned through a timing synchronization transmission module, and the signal transmission interval is dynamically adjusted according to the dust concentration monitoring result. When the dust concentration is higher than the preset critical value, the signal transmission interval is shortened to reduce signal loss caused by dust interference; The superimposed analysis is performed on the current waveform, voltage waveform and three-dimensional vibration signal that have completed timing synchronization, and the intervals with sudden amplitude changes in the three within the same period are extracted to generate a set of transient signals with strictly aligned time stamps as the anti-interference data set.

10. A mine-used circuit fault self-diagnosis system, characterized in that, It includes: An acquisition module that synchronously acquires the current waveform, voltage waveform and three-dimensional vibration signal of the cable through a multi-dimensional sensor group, and generates an anti-interference data set through adaptive noise reduction and timing synchronization transmission in a humid dust environment; An extraction module that dynamically adjusts the waveform intercept window to capture transient waveform segments based on the correlation between the mechanical vibration intensity and the current transient rate in the anti-interference data set, extracts the peak slope change amount from the transient waveform segments of the current waveform, separates the harmonic band energy of the short-circuit spikes from the transient waveform segments of the voltage waveform, and simultaneously performs time-frequency analysis on the three-dimensional vibration signal to generate a set of mechanical damage spectrum features; A correlation module that inputs the peak slope change amount, harmonic band energy and mechanical damage spectrum feature set into a spatio-temporal correlation model, and generates a composite feature vector through spatio-temporal correlation verification of current distortion features and voltage anomaly features; A correction module that dynamically corrects the threshold in the fault feature database based on the environmental interference factor in the composite feature vector, combined with the cable surface temperature and dust data; A matching module that performs multi-dimensional matching calculations on the composite feature vector and the corrected fault mode threshold, and determines the fault diagnosis result according to the correlation strength between the electrical features and the mechanical features in the matching result.

Images