Line insulation inspection and power-on experiment method

Abstract

The invention relates to the technical field of electrical detection, solves the technical problem of monitoring confusion caused by coupling of electrode aging drifting and real insulation fault signals in a frequency domain in a periodic dynamic burst electrical system, and particularly relates to a line insulation inspection and power-on experiment method. Comprising three-phase current signals and line voltage signals; welding periods are divided based on amplitude abrupt change characteristics of the current signals, and each welding period is divided into an arc starting stage, an arc stabilizing stage and an arc quenching stage; and the period stability index is calculated through the stage duration of the welding period, and effective data are screened. According to the method, the fundamental frequency energy ratio, the harmonic pollution index and the phase offset are extracted, the progressive trend of stripping electrode aging is fitted in combination with least square linear fitting, and the frequency domain energy concentration ratio is utilized to accurately distinguish the insulation fault and the electrode aging signal; false fault signals caused by electrode aging are prevented from being misjudged as insulation faults, and unnecessary shutdown maintenance is reduced.

Description

Technical Field

[0001] This invention relates to the field of electrical testing technology, and in particular to methods for circuit insulation inspection and energization testing. Background Technology

[0002] Ensuring the safety and reliability of power lines and equipment is crucial in the operation and maintenance of power systems and electrical equipment. With the continuous development of power systems and the widespread application of electrical equipment, the insulation performance of lines and the energized operating status of equipment directly affect system stability and personnel safety. Therefore, conducting line insulation inspections and energized tests has become an indispensable part of power engineering.

[0003] Line insulation inspection is primarily used to assess the insulation performance of electrical lines, cables, motors, transformers, and other equipment. Good insulation performance can effectively prevent faults such as leakage, short circuits, and breakdowns, thereby avoiding electrical accidents and equipment damage. Energizing tests verify the performance of electrical equipment under actual operating conditions. Through gradual energizing, load testing, and protection device operation testing, a comprehensive understanding of the equipment's operation under energized conditions can be obtained, including changes in parameters such as voltage, current, temperature, and power, as well as the presence of abnormal phenomena such as overheating, unusual noises, or odors. Line insulation inspection and energizing tests are closely related in power systems. Insulation inspection is the prerequisite and foundation for energizing tests; only when insulation performance is ensured can safe energizing tests be conducted. Energizing tests, on the other hand, can further verify the effectiveness of insulation inspection results and discover potential problems that may not have been detected during the insulation inspection under actual operating conditions. The two complement each other, together forming an important guarantee for the safe operation of the power system.

[0004] In industrial automated production scenarios, the reliability of line insulation telemetry and energization testing faces severe challenges, especially in electrical systems exhibiting periodic dynamic and sudden behaviors. In electrical welding production lines, when industrial arc welding robots perform pre-set periodic welding tasks, each welding cycle (e.g., triggered at intervals of several minutes or seconds) involves the electrode momentarily igniting an arc on the workpiece surface, forming a millisecond-level high-current discharge pulse. The arc ignition is accompanied by a strong current surge and voltage fluctuation, followed by a period of stable arc sustaining before the electrode rapidly extinguishes it. This periodic dynamic behavior is inherently sudden: the arc ignition and extinguishing process is rapid, steep, and high-frequency, but its cycle (i.e., welding interval) is controlled by the production rhythm and exhibits regularity. However, due to ablation and wear during long-term use, the arc ignition characteristics (such as surge rise slope and peak duration) of the electrodes gradually change over time. This drift process is slow and continuous, persisting throughout the entire equipment lifecycle.

[0005] Existing line insulation telemetry and energization testing systems, when performing insulation telemetry or energization tests on the power supply circuit of welding production lines, experience indistinguishable coupling effects in the frequency domain due to characteristic drift caused by electrode aging, which in turn couples with actual insulation fault signals. Specifically, this includes: Fundamental frequency band attenuation: Electrode ablation causes attenuation of the fundamental component of the arc initiation current, leading the system to misinterpret it as an increase in line contact resistance; Harmonic pollution: Uneven discharge on the worn surface generates random high-frequency harmonics, whose spectral characteristics are highly similar to those of partial discharge in insulation; Phase shift: The arc initiation delay effect causes phase drift of the electrical signal, which is misinterpreted as an abnormal phase of capacitive leakage current.

[0006] This spectral obfuscation creates a dynamic camouflage effect in the time-frequency joint domain—pseudo-fault signals generated by electrode aging completely overlap with real insulation degradation signals in the following dimensions. This leads the system to misinterpret normal drift signals as insulation faults, triggering shutdowns for maintenance; frequent switching operations actually accelerate line aging, generating real defects. Furthermore, real insulation degradation signals (such as intermittent discharges caused by sheath carbonization) are submerged in drift noise, creating permanent monitoring blind spots. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention provides a method for line insulation inspection and energization testing, which solves the technical problem of monitoring confusion caused by the frequency domain coupling of electrode aging drift and actual insulation fault signals in periodic dynamic sudden electrical systems. This achieves accurate separation of drift signals and fault signals, reduces false alarm rate and improves fault detection rate, thereby extending equipment life.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a method for circuit insulation inspection and energization test, comprising the following steps: Acquire electrical signals from the electrical system, including three-phase current signals and line voltage signals; Welding cycles are divided based on the amplitude change characteristics of current signals, with each welding cycle consisting of an arc initiation stage, an arc stabilization stage, and an arc extinguishing stage. The cycle stability index is calculated based on the duration of each stage, and valid data is then selected. For each welding cycle, a short-time Fourier transform with an adaptive window length is performed on the signal phase to extract the fundamental frequency energy ratio, which reflects energy transfer efficiency, the harmonic contamination index, which detects nonlinear distortion, and the phase offset, which detects phase anomalies, and a time-varying feature matrix is ​​constructed. The electrode aging trend component in the time-varying feature matrix is ​​stripped and the residual sequence containing coupling effect is extracted; the residual sequence is subjected to resonance coupling suppression processing to remove the resonance frequency energy and generate a purified residual sequence. The fault band energy and the full band energy are extracted from the cleanup residual sequence. The energy concentration of the insulation fault is defined as the ratio of the two. The decoupling decision value is calculated to characterize the degree of energy concentration of the insulation fault in the frequency domain. The insulation fault is determined by the decoupling decision value and the fault detection threshold. The drift rate deviation used to trigger electrode aging early warning is calculated based on the electrode aging trend component. Based on the identified insulation fault or electrode aging warning results, corresponding protection measures are triggered.

[0009] Furthermore, the method of dividing the welding cycle based on the amplitude abrupt change characteristics of the current signal includes: When the current in any phase exceeds the preset amplitude threshold, it is marked as the start point of the welding cycle; When the absolute value of the current derivative is less than or equal to the maximum allowable rate of change threshold of the arc state and the minimum duration of the condition is met, it is determined to be the arc initiation and termination point. When the current derivative is less than or equal to the minimum slope threshold of the current drop, it is determined as the arc extinction start point; When the current in any phase is less than or equal to the system noise current level and the absolute value of the derivative of any phase current is less than or equal to the rate of change threshold of zero current, it is determined to be the end point of arc extinction. The welding cycle is divided based on the start or end point of each stage, i.e.: ; In the formula, This is the arc initiation stage; This is the stable arc phase; This is the arc extinction phase; This marks the start point of the welding cycle. The starting and ending points of the arc; This is the starting point for arc extinguishing; This is the point where the arc ends.

[0010] Furthermore, the method for filtering valid data is as follows: Calculate the stage stability index used to quantify the arc energy conversion efficiency for each welding cycle, namely: ; In the formula, This is the stage stability index for the i-th welding cycle; This refers to the duration of the arc initiation phase; Duration of the steady arc phase; This refers to the duration of the arc extinction phase. Calculate the difference between the stage stability index of the welding cycle and the historical benchmark value, and mark the cycles with an absolute difference greater than the dynamic tolerance threshold as abnormal cycles and remove them.

[0011] Furthermore, the method for constructing the time-varying feature matrix is ​​as follows: The final window length is determined by the minimum of the actual duration and theoretical window length of each signal stage. A time-frequency analysis window is then constructed, and an adaptive short-time Fourier transform is performed on the signal stage of each welding cycle. ; In the formula, The input signal represents the original electrical signal at a specific stage in the i-th welding cycle; For integration time; For window functions; For the complex exponent kernel; where u is the imaginary unit; f is the continuous frequency value; and k is the harmonic order index; is a complex value of the time spectrum, representing the amplitude and phase information of a specific signal stage at time t and frequency f in the i-th welding cycle; stage is the identifier of the working stage of the welding process, including the arc initiation stage (arc), the stable arc stabilization stage (stable), and the arc extinguishing stage (ext); The formula for calculating the fundamental frequency energy ratio is: ; In the formula, The fundamental frequency energy ratio; The neighborhood of the fundamental frequency; For a specific signal phase in the i-th welding cycle at time point and frequency Complex values ​​at; where Let q be the center time of the q-th time window. The center frequency of the p-th frequency component; q is the index of the time axis after discretization; p is the index of the frequency axis after discretization. The formula for calculating the harmonic pollution index is: ; In the formula, The harmonic pollution index is K; K is the highest harmonic order set. The fundamental frequency; For a specific signal phase in the i-th welding cycle at time point and the frequency of the kth harmonic order Complex values ​​at; It is the time-domain maximum operator; This represents the maximum amplitude of the k-th harmonic order over the entire time period; This indicates the maximum amplitude of the fundamental frequency over the entire time period; The formula for calculating the phase offset is: ; In the formula, This is the phase offset. Extraction of argument angle for complex vectors; The spectrum of a voltage signal; For the complex conjugate of the spectrum of current; It is a complex conjugate operator; Calculate the feature vectors of the three signal stages in each welding cycle, i.e.: ; This represents the characteristic value of a specific stage in the i-th welding cycle; This represents the fundamental frequency energy ratio at a specific stage in the i-th welding cycle; This represents the harmonic pollution index at a specific stage in the i-th welding cycle; This represents the phase offset at a specific stage within the i-th welding cycle; The three feature vectors in the same welding cycle are vertically connected to form a stage feature vector. Then, the stage feature vectors of N welding cycles are horizontally arranged and transposed to form a time-varying feature matrix.

[0012] Further, the step of stripping the electrode aging trend component from the time-varying feature matrix and extracting the residual sequence includes: The stage eigenvectors in the time-varying feature matrix are fitted using least-squares linear fitting, i.e.: ; In the formula, This represents the feature value of the j-th feature in the i-th welding cycle; Here, represents the drift component; Drift rate; This is the drift intercept term; It is a residual sequence.

[0013] Furthermore, the method for generating the purified residual sequence is as follows: A parametrically excited oscillator model is established to quantify the coupling effect between electrode aging drift and insulation fault impact, namely: ; In the formula, The second derivative of the residual sequence; For damping term; where The damping coefficient; is the linear stiffness term; where b is the linear stiffness coefficient; This is the nonlinear stiffness term; where d is the nonlinear stiffness coefficient. For parameter excitation terms; For external excitation functions; Calculate the control variables used to transform time-domain residual fluctuations into frequency-domain resonance shifts, i.e.: ; In the formula, The residual amplitude; Then, based on the residual amplitude and stiffness coefficient, the resonant frequency points used to provide the target for active suppression are calculated, namely: ; In the formula, The resonant frequency; The residual sequence is analyzed using short-time Fourier transform to extract the amplitude and phase at the resonant frequency; and the amplitude coefficient used to control the strength of the anti-phase cancellation force and the phase angle providing a phase reference for the anti-phase waveform are calculated, i.e.: ; In the formula, The amplitude coefficient; This represents the amplitude of the residual sequence at the resonance frequency. Optimize coefficients based on experience; Phase angle; For phase operators; For frequency domain positioning markers; STFT is a frequency domain representation in complex form; The cleaned residual sequence, which removes energy coupling interference at the resonant frequency and preserves the true drift signal, is calculated using the amplitude coefficient and phase angle. ; In the formula, To purify the residual sequence; Extraction of the real part; For discrete timestamps.

[0014] Furthermore, the calculation of the decoupling decision value used to characterize the degree of energy accumulation of insulation faults in the frequency domain includes: The residual sequence is periodically extended to form an extended sequence used to eliminate spectral leakage; Perform a discrete Fourier transform on the extended sequence, i.e.: ; In the formula, For frequency The power spectral density value at that location; Let M be the extension sequence of the j-th feature in the i-th welding cycle; M is the total length of the extension sequence. Calculate the electromagnetic energy released by the insulation fault within the designated frequency band, i.e.: ; In the formula, Energy in the fault band; Calculate the full-band energy of the signal that reflects the remaining signal after removing the drift component, i.e.: ; In the formula, The total residual energy; The formula for calculating the decoupling decision value is: ; In the formula, is the decoupling decision value of feature j in the i-th welding cycle, which is the core indicator used to quantify the probability of insulation failure.

[0015] Furthermore, the range of the dedicated frequency band is: ,in The lower limit frequency; This is the upper limit frequency.

[0016] Furthermore, the determination of insulation faults through decoupling decision values ​​and fault detection thresholds includes: Extract the historical decision value sequence from the historical normal data and calculate the baseline statistic; Define a drift factor to counteract the progressive characteristic drift caused by electrode aging, namely: ; In the formula, The drift factor; The initial calibration slope; This is the drift sensitivity coefficient; The formula for calculating the fault detection threshold is: ; In the formula, Let be the fault detection threshold for feature j; Confidence factor; The historical decision value of feature j is the average value. The standard deviation of the historical decision values ​​for feature j; when At that time, and when When this occurs, it is determined to be a genuine insulation fault; In the formula, g is the detection cycle number; L is the length of the detection window; This is the preset minimum number of continuous over-limit periods; This is a conditional function, i.e.: .

[0017] Further, the calculation of the drift rate deviation used to trigger the electrode aging early warning based on the electrode aging trend component includes: Calculate the drift rate deviation used to accurately quantify the degree of electrode aging, i.e.: ; In the formula, This refers to the drift rate deviation. The standard error of the drift rate; An electrode aging warning is triggered when the drift rate deviation exceeds a preset critical threshold.

[0018] By employing the above technical solution, the present invention provides a method for circuit insulation inspection and energization testing, which has at least the following beneficial effects: 1. This invention extracts the fundamental frequency energy ratio, harmonic pollution index, and phase offset through short-time Fourier transform (STFT), and combines it with least-squares linear fitting to remove the gradual trend of electrode aging. It accurately distinguishes insulation faults from electrode aging signals by utilizing frequency domain energy concentration; it avoids false fault signals caused by electrode aging being misjudged as insulation faults, reducing unnecessary downtime for maintenance; and it enhances the detection sensitivity of high-frequency insulation faults (such as partial discharge and carbonization breakdown) through power spectrum analysis of residual sequences, preventing real faults from being drowned out by noise.

[0019] 2. This invention dynamically adjusts the STFT window length according to the arc initiation, arc stabilization, and arc extinguishing stages of the welding cycle, and calculates the fundamental frequency, harmonics, and phase characteristics of each stage to construct a time-varying feature matrix; it captures high-frequency details in the transient process (such as arc initiation) and analyzes low-frequency drift in the steady-state process (such as arc stabilization), thereby improving the accuracy of feature extraction; by dividing the stages by current derivative, the features are strictly correlated with the physical process of the arc, avoiding time-series drift interference caused by electrode aging.

[0020] 3. This invention models the residual sequence as a nonlinear oscillator system and actively suppresses the energy coupling at the resonance frequency point to eliminate the spectral confusion between electrode aging and insulation faults. The purified residual signal makes the fault characteristics purer and improves the reliability of the decoupling decision value. Moreover, after suppressing resonance interference, the adaptive threshold can more accurately distinguish between insulation faults and electrode aging, reducing the false operation of the protection device. Attached Figure Description

[0021] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments of this application and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the experimental method flow in an embodiment of the present invention. Detailed Implementation

[0022] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. This will allow for a full understanding of how the present application uses technical means to solve technical problems and achieve technical effects, and to facilitate its implementation.

[0023] This embodiment proposes a method for line insulation inspection and energization testing. Through dynamic spectrum decoupling, in periodic sudden electrical systems, it: accurately separates electrode aging drift signals from actual insulation fault signals; quantitatively diagnoses insulation status (fundamental frequency attenuation / harmonic pollution / phase shift), reducing false alarm rates; and through real-time feedback control, triggers insulation fault protection or adaptively adjusts equipment parameters to delay electrode aging, reducing unnecessary downtime for maintenance and improving system safety and operational efficiency. Figure 1 As shown, the method includes the following steps: Acquire electrical signals from the power supply circuit of the welding robot, including three-phase current signals (including phase A current). Phase A current C-phase current ) and line voltage signals (including voltage between lines A and B) Voltage between BC lines CA line voltage Three-phase currents together constitute the energy transmission carrier of an electrical system. In arc welding scenarios, these currents include: the surge pulse (microsecond level) at the moment of arc ignition, the steady-state current during the arc stabilization phase, and the decaying current during the arc extinguishing process. Line voltage reflects the potential difference between phases, and its characteristics include: voltage drop at arc ignition (due to a sudden increase in load), voltage stability during arc stabilization, and voltage recovery during arc extinguishing. The asymmetry of the line voltage signal is used to detect phase-to-phase insulation degradation; if the insulation between phases A and B is damaged, abnormal fluctuations in the voltage between the A and B lines will occur. Combined with the three-phase current signals to calculate instantaneous power, the electrode state can be revealed during the arc ignition phase through the differential relationship between current surge and voltage drop. Simultaneously, by comparing the amplitude and phase difference of the three-phase current signals, the faulty phase can be located (e.g., abnormal increase in phase C current when phase C is leaking). A high sampling rate ensures the waveform integrity of millisecond-level transient processes (such as arc ignition).

[0024] It is important to note that the sampling rate for acquiring the three-phase current and line voltage signals must satisfy the Nyquist-Shannon sampling theorem to ensure that the system can completely capture the high-frequency transient components generated during the arc initiation / exit process of the electrodes; that is... ;in, Sampling rate; This represents the highest frequency component that may occur in the system (e.g., higher harmonics of electric arc discharge). All signals are sampled using the same clock source, and the clock synchronization deviation of the sampling satisfies: ;in, This is due to clock synchronization deviation; This refers to the duration of the shortest transient event in the system (e.g., the arc-starting pulse width). Forcing all signal channels (current / voltage) to use the same clock source for sampling, ensuring the time difference between channels is less than 10% of the shortest transient event (e.g., the arc-starting pulse width), eliminates phase misalignment between current and voltage signals, thereby ensuring the reliability of subsequent data and avoiding distortion in subsequent correlation analysis. Furthermore, it is necessary to continuously acquire data for N complete welding cycles to ensure sufficient coverage of drift change information.

[0025] Each welding cycle is defined as the complete process from the start of one arc ignition to the start of the next arc ignition. The continuous waveform is transformed into a discrete periodic sequence, and a time baseline coordinate system is established.

[0026] Utilizing the abrupt amplitude change characteristic of current signals, the welding cycle start point is marked when the current in any phase jumps from the background noise level to a significant amplitude (i.e., exceeding a preset amplitude threshold). Where i is the welding cycle number, For the time domain range of each welding cycle, The purpose of its right-open interval design is to avoid the overlap of cycle endpoints (to prevent time points from being repeatedly assigned to two cycles).

[0027] The amplitude threshold is used to distinguish between operating current and idle noise. When the current of any phase exceeds the amplitude threshold, the equipment is considered to have entered an effective operating cycle. This is achieved by collecting background noise data of the three-phase current signals during the equipment's idle state, calculating the mean and standard deviation of the noise for each phase, and then setting the threshold according to confidence principles.

[0028] Each welding cycle is divided into three stages: the arc initiation stage, where the current rapidly rises from zero to a stable value; the arc stabilization stage, where the current remains relatively stable; and the arc extinguishing stage, where the current decays from a stable value to zero. Based on the physical properties of the current derivative, the start or end point of each stage is determined. The specific determination method is as follows: At the absolute value of the current derivative And continue When the arc begins or ends, it is determined to be the starting and ending point. That is, the starting point of the stable arc. Among them, is the instantaneous rate of change of current; x is the index of any phase; Let x be the phase current; The maximum allowable rate of change threshold for the stable arc state is obtained through statistical analysis of historical stable arc phase data, and the 95th percentile value is taken. The minimum duration for which the condition is met is set to be greater than the maximum duration of system interference.

[0029] In current derivative When the arc is extinguished, it is determined to be the starting point of the arc extinction. .in, The minimum slope threshold for current decrease is calibrated using the electrode response time parameter and measured current change rate data from the welding equipment technical manual.

[0030] In any phase current and When the arc is extinguished, it is determined to be the end point of arc extinction. .in, The system noise current level; The threshold for the rate of change of zero current is obtained by taking the 99% confidence upper limit of the rate of change of background noise.

[0031] By dividing each welding cycle into stages, the timing drift problem caused by electrode aging can be solved, enabling physically driven adaptive segmentation of the working cycle. This provides phase-aligned and feature-pure analysis units for subsequent electrode aging modeling and insulation fault detection. The current derivative is the essential physical characterization of the arc's dynamic behavior; arc initiation / extinction is a sudden energy state transition process, i.e.: Where E represents the arc energy. Therefore, the current derivative directly reflects the plasma ionization / deionization rate of the gap between the electrode and the workpiece, and is more state-resolved than the current value. This allows the scheme to fundamentally eliminate temporal drift contamination caused by electrode aging through physical-driven segmentation using the current derivative, enhances the signal-to-noise ratio of insulation fault feature extraction, lays a reliable foundation for subsequent stable arc segment drift modeling and step residual fault detection, and ultimately achieves a strict correspondence between the segmentation rules and the physical essence of the arc.

[0032] Based on the start or end point of each stage, the welding cycle is divided into: Arc initiation stage ,Right now: ; steady arc phase ,Right now: ; Arc Extinguishing Phase ,Right now: ; In the formula, X represents any phase current; X is the index of any term current.

[0033] By dividing the welding cycle into stages, precise isolation and feature purification of the physical processes are achieved, providing a data foundation with clear physical meaning and homogeneous features for subsequent electrode aging modeling and insulation fault diagnosis, thus completely solving the feature contamination problem caused by mixed original signals. In particular, during fault location, high-frequency insulation defects can be highlighted in the arc initiation stage, while electrode aging drift can be accurately modeled in the arc stabilization stage, eliminating interference from irrelevant stages (such as arc extinguishing oscillations not affecting arc stabilization analysis).

[0034] Calculate the stage stability index used to quantify the arc energy conversion efficiency for each welding cycle, namely: ; In the formula, , which is the stage stability index (dimensionless) for the i-th welding cycle; The arc initiation phase duration represents the time required for the arc to form and reflects the ionization efficiency. This refers to the duration of the arc stabilization phase, representing the effective welding time. The duration of the arc extinction phase reflects the energy dissipation time and is related to the deionization rate.

[0035] The difference between the stage stability index of the welding cycle and the historical benchmark value is compared with the dynamic tolerance threshold. Cycles with an absolute difference greater than the dynamic tolerance threshold are marked as abnormal cycles and removed. ; In the formula, As a historical baseline, the latest value is extracted. The average value of the stage stability index for each normal welding cycle is obtained, namely: ; The dynamic tolerance threshold is set by calculating the standard deviation of the stage stability index during a normal welding cycle and based on the three sigma principle; that is: ; Simultaneously, the dynamic tolerance threshold is dynamically adjusted based on the dispersion coefficient to achieve intelligent filtering of external interference and adaptive tracking of electrode aging, i.e.: ; In the formula, This is the dynamic tolerance threshold for the current welding cycle, used to determine the boundary line of abnormal welding cycles in real time. This is the dynamic tolerance threshold of the previous welding cycle, used to provide an adjustment benchmark and ensure continuity; This is an empirical coefficient, obtained based on verification of historical welding data; The coefficient of variation is a normalization metric (dimensionless) used to measure the dispersion of data. The standard deviation of the phase stability index is used to quantify the normal fluctuation range, i.e.: ; By detecting anomalies in stage stability indicators, atypical welding cycles caused by external interference are eliminated, ensuring the purity of data and the reliability of equipment status characterization in subsequent electrode aging and fault analysis.

[0036] An adaptive window-length short-time Fourier transform is applied to the signal phase of each welding cycle to achieve high-resolution time-frequency domain decomposition and multi-component separation of non-stationary signals, providing optimized time-frequency representation for feature extraction. This allows for an optimal balance between time-frequency resolution and different working stages (arc initiation / arc stabilization / arc extinction), capturing details in the transient process and analyzing spectral characteristics in the steady-state process. The specific operation method is as follows: The theoretical window length is calculated based on the target frequency resolution. The final window length (i.e., the adaptive window length) is then determined by the minimum of the actual duration of each signal stage and the theoretical window length. ; In the formula, The window length determines the length of each time window in the short-time Fourier transform, affecting the time resolution and frequency resolution of time-frequency analysis. The actual duration of the current signal stage; stage is the identifier of the working stage of the welding process, including: arc initiation stage (arc), arc stabilization stage (stable), and arc extinguishing stage (ext); Adjustment factor, used to correct for differences in main lobe width among different window types; This refers to the resolution frequency, which is the set lower limit of the resolution.

[0037] A time-frequency analysis window is constructed using the Blackman-Harris window function, which achieves sidelobe suppression exceeding 90 dB through a specific combination of coefficients; that is: ; In the formula, is a window function, representing the amplitude of the window function at time point t, with a value range of [0,1], used to control the time-domain weight distribution of the signal stage; is the cosine function; k is the index of the harmonic order; is the amplitude weight of the k-th harmonic, used to maximize sidelobe suppression.

[0038] For each signal phase of the welding cycle, an adaptive window-length short-time Fourier transform is performed, i.e.: ; In the formula, The input signal represents the original electrical signal at a specific stage in the i-th welding cycle; This is the integration time variable, used to iterate through all time points within the time window and perform integration calculations; It is a window function used to suppress spectral leakage and enhance signal weight in the central region; Indicates the time offset; The complex exponential kernel is used to project a time-domain signal onto the frequency domain and extract frequency components; where u is the imaginary unit and f is a continuous frequency value used to define the frequency points for analysis. For the time-frequency complex value, it represents the amplitude and phase information of a specific signal stage at time t and frequency f in the i-th welding cycle, and is essentially a two-dimensional complex matrix.

[0039] The fundamental frequency energy ratio, reflecting energy transfer efficiency, the harmonic contamination index, used to detect nonlinear distortion, and the phase shift, used to detect phase anomalies, are calculated using time-spectrum complex numerical methods; that is: ; In the formula, This is the fundamental frequency energy ratio, used to quantify the proportion of the fundamental component in the total energy; electrode ablation will cause this value to decrease (increase contact resistance), while insulation degradation has little effect on it; This is the neighborhood of the fundamental frequency, used to locate the concentrated band of fundamental frequency energy. For a specific signal phase in the i-th welding cycle at time point and frequency Complex values ​​at; where Let q be the center time of the q-th time window. The center frequency of the p-th frequency component; q is the index of the time axis after discretization; p is the index of the frequency axis after discretization. ; In the formula, The harmonic pollution index is used to quantify the intensity of harmonic components relative to the fundamental frequency; electrode wear (uniform surface discharge) and partial discharge of insulation will both cause this value to increase; k is the index of the harmonic order; K is the set highest harmonic order; The fundamental frequency; For a specific signal phase in the i-th welding cycle at time point and the frequency of the kth harmonic order Complex values ​​at; It is a time-domain maximum value operator used for peak values ​​(not average values) across the entire time period, thereby highlighting transient pulses; This represents the maximum amplitude of the k-th harmonic order over the entire time period; This represents the maximum amplitude of the fundamental frequency over the entire time period (i.e., the normalized reference).

[0040] ; In the formula, This is the phase offset, used to extract the phase difference between voltage and current at the fundamental frequency; electrode arcing delay will cause a linear drift in this value, while capacitive leakage current will cause a step change. This is used for complex vector argument extraction, which is then used to output the phase offset. The voltage signal's frequency spectrum reflects the complex amplitude of the voltage at the fundamental frequency and is used to provide a reference for phase measurement. It is the complex conjugate of the current frequency spectrum, reflecting the complex amplitude conjugate of the current at the fundamental frequency, and is used to construct a phase difference measurement channel; This is a complex conjugate operator.

[0041] By quantifying the proportion of fundamental frequency energy, the increased contact resistance caused by electrode ablation is sensitively captured, while avoiding interference from insulation degradation. Secondly, by uniformly quantifying nonlinear distortion through the relative values ​​of harmonic intensity, the spectral characteristics of electrode surface discharge and insulation partial discharge are synchronously responded to. The dynamic characteristics of fundamental frequency phase are accurately extracted, effectively distinguishing between the gradual time delay of electrode aging and the step leakage characteristics of insulation faults.

[0042] Calculate the feature vectors of the three signal stages in each welding cycle, i.e.: ; This represents the characteristic value of a specific stage in the i-th welding cycle; This represents the fundamental frequency energy ratio at a specific stage in the i-th welding cycle; This represents the harmonic pollution index at a specific stage in the i-th welding cycle; This represents the phase offset at a specific stage in the i-th welding cycle.

[0043] The three feature vectors from the same welding cycle are vertically concatenated to form a stage feature vector, i.e.: ; In the formula, is the stage feature vector of the i-th welding cycle; ∥ is the vertical concatenation operator (representing the vertical connection of vectors); R represents a 9-dimensional column vector space (3 stages × 3 features / stage); R represents the set of real numbers.

[0044] Then, the stage feature vectors of N welding cycles are arranged horizontally and transposed to form a time-varying feature matrix, i.e.: ; In the formula, represents the time-varying feature matrix; / is the horizontal concatenation operator (indicating horizontal concatenation of vectors); T represents matrix transpose; express 3D real matrix format.

[0045] By integrating fragmented features from multiple stages and cycles into structured spatiotemporal data, a standardized matrix is ​​constructed that simultaneously contains the cyclic evolution time sequence (row direction) and physical state description (column direction), providing directly computable mathematical input for subsequent drift-fault decoupling. Through temporal decoupling, N welding cycles are arranged in the row direction to explicitly preserve the temporal evolution trajectory of electrode aging / insulation degradation. Through spatial decoupling, nine features are arranged in the column direction to achieve: arc initiation features (columns 1-3) specifically for monitoring electrode surface ablation; arc stabilization features (columns 4-6) specifically for tracking line impedance attenuation; and arc extinguishing features (columns 7-9) specifically for capturing insulation degradation.

[0046] Least-squares linear fitting is performed on the stage eigenvectors in the time-varying feature matrix to remove the progressive electrode aging trend component; and the residual sequence is extracted to eliminate drift interference such as fundamental frequency attenuation and harmonic pollution. The principle is that electrode aging is a progressive process, and its changes exhibit a linear trend in the time domain; while insulation faults are sudden events, manifesting as instantaneous abnormal fluctuations in eigenvalues. Therefore, when the two are superimposed, the superposition principle applies. ; In the formula, This represents the feature value of the j-th feature in the i-th welding cycle, used as the input to the decoupling algorithm, and contains a mixed signal of drift and fault. Here, represents the drift component; The drift rate represents the slope of the linear trend of characteristic j changing with time (welding cycle), describing the average rate at which the characteristic value changes in each cycle due to electrode aging, and is used to capture the progressive degradation trend. The drift intercept term reflects the initial aging offset, which is the eigenvalue offset that already exists in the first cycle (originating from historical aging), and is used to eliminate the interference of historical equipment degradation on the current detection. Let be the residual sequence, representing the portion of the eigenvalue of feature j remaining after removing the drift component in the i-th welding cycle; this portion includes abnormal fluctuations caused by insulation faults (sudden faults), as well as other random noise; ideally, it should be zero or very small if there are no insulation faults; if there are insulation faults, it will deviate significantly from zero.

[0047] The residual sequence is periodically extended by adding MN zeros to the end of the residual sequence to make the sequence length meet the requirements of Fast Fourier Transform (FFT), forming an extended sequence used to eliminate spectral leakage. Here, M is the smallest integer power of 2 greater than or equal to N, i.e.: ; When performing a Fourier transform directly on a finite-length sequence, it's equivalent to truncating an infinitely long signal with a rectangular window. This time-domain truncation leads to sinc function sidelobes in the frequency domain, causing energy leakage. Zero-padding is equivalent to performing higher-density frequency domain sampling on the original sequence, which manifests as interpolation smoothing of the original spectrum in the frequency domain. The sidelobe energy is compressed into the high-frequency region (away from the fault characteristic frequency band). This prevents the high-frequency pulse energy of insulation faults from being overwhelmed by low-frequency drifting sidelobes.

[0048] The extended sequence is processed by Discrete Fourier Transform, which transforms the time-domain residual into a frequency-domain energy distribution, i.e.: ; In the formula, For frequency The power spectral density value at that point, i.e. the energy intensity of the extended sequence, is expressed in W / Hz; Let M be the extension sequence of the j-th feature in the i-th welding cycle; M is the total length of the extension sequence.

[0049] Insulation faults (such as partial discharge or sheath carbonization) can generate electromagnetic pulses of specific frequencies, which manifest in the frequency domain as... Energy accumulation within the frequency band. Among them, The lower limit frequency is the lowest effective frequency of the insulation fault signal in the spectrum. Energy below this frequency mainly comes from electrode aging noise (such as fundamental frequency attenuation) and is unrelated to the actual fault. The upper limit frequency is the highest effective frequency of the insulation fault signal in the frequency spectrum. Energy above this frequency belongs to random noise (such as switching transients) and no longer contains fault characteristics. The insulation fault energy concentration is defined as the ratio of the fault band energy to the total frequency band energy. The energy concentration is calculated, and a decoupling decision value is output to characterize the degree of energy concentration of the insulation fault in the frequency domain, i.e.: ; In the formula, , is the decoupling decision value of feature j in the i-th welding cycle, which is the core indicator used to quantify the probability of insulation failure; The fault band energy indicates the insulation fault occurring in a specific frequency band. The electromagnetic energy released internally, namely: ; The total residual energy reflects the full-band energy of the signal remaining after removing the drift component, i.e.: ; By decoupling the spectral coupling effect of electrode aging (low-frequency drift) and insulation fault (high-frequency burst), the problem of overlap and confusion between the two in the frequency domain is solved, allowing the insulation fault characteristics to be independently presented in the residual power spectrum, thus eliminating the interference of spurious fault signals caused by electrode aging. Furthermore, the insulation fault can be abstracted into a quantifiable index through energy concentration.

[0050] Extract the historical decision value sequence from the historical normal data and calculate the benchmark statistics, including the historical mean and historical standard deviation, i.e.: ; In the formula, Let E be the average of the historical decision values ​​for feature j, reflecting the central trend of feature j; E is the number of historical data points, which must satisfy... To ensure statistical significance; The decoupling decision value of feature j in the i-th welding cycle of the historical decision value sequence; The standard deviation of the historical decision values ​​of feature j reflects the degree of dispersion of feature j.

[0051] Define a drift factor to counteract the progressive characteristic drift caused by electrode aging, namely: ; In the formula, This is a drift factor used to address the interference of gradual feature drift caused by electrode aging on the fault detection threshold, thereby achieving dynamic adaptive thresholding. This is the initial calibration slope, i.e., the reference drift rate under healthy equipment conditions; The drift sensitivity coefficient is used to adjust the intensity of compensation and is determined based on historical experimental data.

[0052] The fault detection boundary for the adaptive electrode aging process is calculated based on the drift factor, namely: ; In the formula, The fault detection threshold for feature j can automatically adapt to feature drift caused by electrode aging, realize the synchronous evolution of the threshold and equipment status, and thus eliminate the interference of progressive electrode aging drift on the detection system without reducing the sensitivity of insulation fault detection. The confidence factor is used to adjust the threshold boundary width, which is determined based on experimental data.

[0053] The fundamental difference between electrode aging and insulation failure lies in their variation characteristics and time scale. Electrode ablation causes electrical characteristics (such as arc ignition speed) to deteriorate linearly over time; in terms of variation pattern, the drift rate continuously deviates from the initial value, with small changes but a fixed direction (unidirectional decay); in terms of time scale, it spans hundreds to thousands of welding cycles. Insulation failure, on the other hand, is caused by the breakdown of the insulating material, triggering a transient abnormal discharge; in terms of variation pattern, the decoupling judgment value increases sharply at the moment of failure, exhibiting a pulse-like pattern; in terms of time scale, it occurs within one to several cycles. By comparing the decoupling judgment value with the fault detection threshold, through initial screening for amplitude exceeding limits and continuous verification and re-judgment, sudden insulation failures can be accurately detected and transient interference eliminated, achieving reliable alarm. In short: when When this occurs, it indicates an abnormal energy accumulation, meaning the amplitude exceeds the limit; And when This indicates a persistent anomaly, requiring continuous verification. In the formula, g is the detection cycle number, g=i represents the current cycle, g=i-1 represents the previous welding cycle, g=iL represents the earliest welding cycle within the detection window; L is the length of the detection window. This is the preset minimum number of continuous over-limit periods; This is a conditional function, i.e.: ; If both of the above conditions are met simultaneously, it can be determined as a genuine insulation fault. If only the amplitude exceeds the limit but is not sustained, it can be determined as transient interference. Through a dual verification mechanism, genuine insulation faults (such as line breakdown, partial discharge, etc.) can be accurately identified, and genuine faults (continuous energy anomalies) can be distinguished from transient interference (voltage flicker, electromagnetic noise, etc.). Furthermore, by coordinating multi-dimensional features, the fundamental frequency, harmonics, and phase characteristics can be independently judged, avoiding misjudgment based on a single feature.

[0054] Calculate the drift rate deviation used to accurately quantify the degree of electrode aging, i.e.: ; In the formula, The drift rate deviation was quantified to determine the statistical significance of electrode aging, eliminate measurement noise interference, and provide an objective and reliable decision-making basis for aging early warning. This represents the standard error of the drift rate.

[0055] When the drift rate deviation exceeds a preset critical threshold, an electrode aging warning is triggered, enabling precise monitoring and proactive maintenance decisions for electrode aging.

[0056] Finally, based on the output decision result (actual insulation fault or electrode aging warning), the corresponding decision-making measures are triggered: in the event of an insulation fault, the protection device is tripped, and fault phase location information is output. In the event of electrode aging, the energizing test parameters are adjusted (e.g., reducing the test current by 20%) to extend the electrode life.

[0057] It should be noted that characteristic drift and insulation faults exhibit resonant coupling in the residual sequence spectrum (0.1–10 Hz band), causing the fault signal energy to be masked and affecting the accuracy of spectral decoupling. This is because nonlinear drift caused by electrode aging (such as abrupt changes in ablation rate) and the impulsive signal of insulation faults mutually excite each other within the frequency band, forming parametric excitation-type resonance, which masks the fault characteristics with the harmonic energy of the drift. Therefore, it is necessary to actively suppress resonant frequencies, purify the residual signal, and eliminate coupling interference to improve the signal-to-noise ratio and detectability of fault characteristics in the decoupling decision function. Avoiding resonant coupling amplifies the risk of misjudgment, leading to missed detection of real faults or false alarms of normal drift. The method for purifying the residual signal is as follows: The coupling effect of electrode aging drift and insulation fault impact in the residual sequence is transformed into a quantifiable physical system, namely: ; In the formula, The second derivative of the residual sequence, i.e., the acceleration term, is used to quantify the abrupt change intensity of the residual sequence and reflect the degree of drastic signal changes caused by electrode aging drift or insulation faults. This is the damping term, used to suppress high-frequency noise energy and prevent system oscillation divergence; where... The damping coefficient is obtained by multiplying the harmonic pollution index by a preset calibration constant (a dimensionless proportionality coefficient determined in advance through equipment aging experiments). Its core function is to adaptively suppress high-frequency oscillation noise to stabilize the dynamic response of the system. is the linear stiffness term used to constrain the linear recovery characteristics of baseline drift; where b is the linear stiffness coefficient used to control the linear restoring force of the oscillator near the equilibrium position, obtained through recursive least squares identification of the residual sequence; The term is a nonlinear stiffness term used to model the gradual drift abrupt changes caused by electrode ablation (such as accelerated wear); where d is a nonlinear stiffness coefficient used to control the stiffness enhancement effect of the oscillator under large displacement, and is identified and obtained simultaneously with the linear stiffness coefficient. As a parameter excitation term, the energy impact of insulation faults (such as partial discharge and carbonization breakdown) is converted into parameter excitation in the oscillator system; The external excitation function describes the transient impact characteristics of insulation faults. This formula represents the external excitation formed by the superposition of Y fault-induced insulation pulses at time t. Each pulse is a Gaussian envelope-modulated cosine wave (i.e., a Gabor wavelet). The transient electromagnetic pulse sequence generated by the insulation fault is simulated. These pulses are localized in the time domain and selective in the frequency domain. The pulses are dynamically generated by analyzing the power spectral density (PSD) and short-time Fourier transform results of the residual sequence.

[0058] Calculate the control variables used to transform time-domain residual fluctuations into frequency-domain resonance shifts, i.e.: ; In the formula, The residual amplitude, i.e., the root mean square of the residual sequence, quantifies the overall amplitude of the mixed oscillation of electrode aging drift and insulation fault.

[0059] Then, based on the residual amplitude and stiffness coefficient, the resonant frequency points used to provide the target for active suppression are calculated, namely: ; In the formula, This is the resonant frequency point, used to analyze the coupling band between electrode aging and insulation faults.

[0060] The residual sequence is analyzed using short-time Fourier transform to extract the amplitude and phase at the resonant frequency; and the amplitude coefficient used to control the strength of the anti-phase cancellation force and the phase angle providing a phase reference for the anti-phase waveform are calculated, i.e.: ; In the formula, The amplitude coefficient; This represents the amplitude of the residual sequence at the resonance frequency. Optimize coefficients based on experience; Phase angle; It is a phase operator used to capture the rotational phase information of a signal; The STFT is a frequency domain location marker that emphasizes taking values ​​only at the resonant frequency point; the STFT is a complex form of frequency domain representation that provides complete frequency domain information for both the real and imaginary parts.

[0061] The cleaned residual sequence, which removes energy coupling interference at the resonant frequency and preserves the true drift signal, is calculated using the amplitude coefficient and phase angle. ; In the formula, To purify the residual sequence and to eliminate the pure electrode drift signal after resonant coupling; Real part extraction is used to convert complex spectrum into real signal waveform; For discrete timestamps.

[0062] Subsequent processing based on the purified residual sequence utilizes a nonlinear resonance suppression mechanism to fundamentally solve the deep coupling problem between electrode aging drift and insulation faults in the critical frequency band. This achieves more thorough feature separation, significantly improving the purity of fault features in the residual signal and providing interference-free input for subsequent spectrum analysis. Furthermore, the purified residual sequence greatly reduces the ambiguity of the decoupling decision value, enabling adaptive threshold decisions to clearly distinguish between insulation faults and electrode aging, thus avoiding false triggering of protection devices.

[0063] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0064] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. Since the above embodiments are substantially similar to the method embodiments, their descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0065] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for circuit insulation inspection and energization test, characterized in that, Includes the following steps: Acquire electrical signals from the electrical system, including three-phase current signals and line voltage signals; Welding cycles are divided based on the amplitude change characteristics of current signals, with each welding cycle consisting of an arc initiation stage, an arc stabilization stage, and an arc extinguishing stage. The cycle stability index is calculated based on the duration of each stage, and valid data is then selected. For each welding cycle, a short-time Fourier transform with an adaptive window length is performed on the signal phase to extract the fundamental frequency energy ratio, which reflects energy transfer efficiency, the harmonic contamination index, which detects nonlinear distortion, and the phase offset, which detects phase anomalies, and a time-varying feature matrix is ​​constructed. The electrode aging trend component in the time-varying feature matrix is ​​stripped and the residual sequence containing coupling effects is extracted; The residual sequence is subjected to resonant coupling suppression processing to remove the resonant frequency energy and generate a purified residual sequence. The fault band energy and the full band energy are extracted from the cleanup residual sequence. The energy concentration of the insulation fault is defined as the ratio of the two. The decoupling decision value used to characterize the degree of energy concentration of the insulation fault in the frequency domain is calculated. Insulation faults are determined by decoupling decision values ​​and fault detection thresholds; The drift rate deviation used to trigger electrode aging early warning is calculated based on the electrode aging trend component. Based on the identified insulation fault or electrode aging warning results, corresponding protection measures are triggered.

2. The method for circuit insulation inspection and energization test according to claim 1, characterized in that, The method of dividing the welding cycle based on the amplitude change characteristics of the current signal includes: When the current in any phase exceeds the preset amplitude threshold, it is marked as the start point of the welding cycle; When the absolute value of the current derivative is less than or equal to the maximum allowable rate of change threshold of the arc state and the minimum duration of the condition is met, it is determined to be the arc initiation and termination point. When the current derivative is less than or equal to the minimum slope threshold of the current drop, it is determined as the arc extinction start point; When the current in any phase is less than or equal to the system noise current level and the absolute value of the derivative of any phase current is less than or equal to the rate of change threshold of zero current, it is determined to be the end point of arc extinction. The welding cycle is divided based on the start or end point of each stage, i.e.: ； In the formula, This is the arc initiation stage; This is the stable arc phase; This is the arc extinction phase; This marks the start point of the welding cycle. The starting and ending points of the arc; This is the starting point for arc extinguishing; This is the point where the arc ends.

3. The method for circuit insulation inspection and energization test according to claim 2, characterized in that, The method for filtering valid data is as follows: Calculate the stage stability index used to quantify the arc energy conversion efficiency for each welding cycle, namely: ； In the formula, This is the stage stability index for the i-th welding cycle; This refers to the duration of the arc initiation phase; Duration of the steady arc phase; This refers to the duration of the arc extinction phase. Calculate the difference between the stage stability index of the welding cycle and the historical benchmark value, and mark the cycles with an absolute difference greater than the dynamic tolerance threshold as abnormal cycles and remove them.

4. The method for circuit insulation inspection and energization test according to claim 1, characterized in that, The method for constructing the time-varying feature matrix is ​​as follows: The final window length is determined by the minimum of the actual duration and theoretical window length of each signal stage. A time-frequency analysis window is then constructed, and an adaptive short-time Fourier transform is performed on the signal stage of each welding cycle. ； In the formula, The input signal represents the original electrical signal at a specific stage in the i-th welding cycle; For integration time; For window functions; For the complex exponent kernel; where u is the imaginary unit; f is the continuous frequency value; and k is the harmonic order index; is a complex value of the time spectrum, representing the amplitude and phase information of a specific signal stage at time t and frequency f in the i-th welding cycle; stage is the identifier of the working stage of the welding process, including the arc initiation stage (arc), the stable arc stabilization stage (stable), and the arc extinguishing stage (ext); The formula for calculating the fundamental frequency energy ratio is: ； In the formula, The fundamental frequency energy ratio; The neighborhood of the fundamental frequency; For a specific signal phase in the i-th welding cycle at time point and frequency Complex values ​​at; where Let q be the center time of the q-th time window. The center frequency of the p-th frequency component; q is the index of the time axis after discretization; p is the index of the frequency axis after discretization. The formula for calculating the harmonic pollution index is: ； In the formula, The harmonic pollution index is K; K is the highest harmonic order set. The fundamental frequency; For a specific signal phase in the i-th welding cycle at time point and the frequency of the kth harmonic order Complex values ​​at; It is the time-domain maximum operator; This represents the maximum amplitude of the k-th harmonic order over the entire time period; This indicates the maximum amplitude of the fundamental frequency over the entire time period; The formula for calculating the phase offset is: ； In the formula, This is the phase offset. Extraction of argument angle for complex vectors; The spectrum of a voltage signal; For the complex conjugate of the spectrum of current; It is a complex conjugate operator; Calculate the feature vectors of the three signal stages in each welding cycle, i.e.: ； This represents the characteristic value of a specific stage in the i-th welding cycle; This represents the fundamental frequency energy ratio at a specific stage in the i-th welding cycle; This represents the harmonic pollution index at a specific stage in the i-th welding cycle; This represents the phase offset at a specific stage within the i-th welding cycle; The three feature vectors in the same welding cycle are vertically connected to form a stage feature vector. Then, the stage feature vectors of N welding cycles are horizontally arranged and transposed to form a time-varying feature matrix.

5. The method for circuit insulation inspection and energization test according to claim 1, characterized in that, The process of stripping the electrode aging trend component from the time-varying feature matrix and extracting the residual sequence includes: The stage eigenvectors in the time-varying feature matrix are fitted using least-squares linear fitting, i.e.: ； In the formula, This represents the feature value of the j-th feature in the i-th welding cycle; Here, represents the drift component; This refers to the drift rate; This is the drift intercept term; It is a residual sequence.

6. The method for circuit insulation inspection and energization test according to claim 5, characterized in that, The method for generating the purified residual sequence is as follows: A parametrically excited oscillator model is established to quantify the coupling effect between electrode aging drift and insulation fault impact, namely: ； In the formula, The second derivative of the residual sequence; For damping term; where The damping coefficient; is the linear stiffness term; where b is the linear stiffness coefficient; This is a nonlinear stiffness term; Where d is the nonlinear stiffness coefficient; For parameter excitation terms; For external excitation functions; Calculate the control variables used to transform time-domain residual fluctuations into frequency-domain resonance shifts, i.e.: ； In the formula, The residual amplitude; Then, based on the residual amplitude and stiffness coefficient, the resonant frequency points used to provide the target for active suppression are calculated, namely: ； In the formula, The resonant frequency; The amplitude and phase at the resonance frequency point are extracted by analyzing the residual sequence using short-time Fourier transform. And calculate the amplitude coefficient used to control the strength of the anti-phase cancellation force and the phase angle that provides a phase reference for the anti-phase waveform, that is: ； In the formula, The amplitude coefficient; This represents the amplitude of the residual sequence at the resonance frequency. Optimize coefficients based on experience; The phase angle; For phase operators; For frequency domain positioning markers; STFT is a frequency domain representation in complex form; The cleaned residual sequence, which removes energy coupling interference at the resonant frequency and preserves the true drift signal, is calculated using the amplitude coefficient and phase angle. ； In the formula, To purify the residual sequence; Extraction of the real part; For discrete timestamps.

7. The method for circuit insulation inspection and energization test according to claim 6, characterized in that, The calculation of the decoupling decision value, used to characterize the energy accumulation degree of insulation faults in the frequency domain, includes: The residual sequence is periodically extended to form an extended sequence used to eliminate spectral leakage; Perform a discrete Fourier transform on the extended sequence, i.e.: ； In the formula, For frequency The power spectral density value at that location; Let M be the extension sequence of the j-th feature in the i-th welding cycle; M is the total length of the extension sequence. Calculate the electromagnetic energy released by the insulation fault within the designated frequency band, i.e.: ； In the formula, Energy in the fault band; Calculate the full-band energy of the signal that reflects the remaining signal after removing the drift component, i.e.: ； In the formula, The total residual energy; The formula for calculating the decoupling decision value is: ； In the formula, is the decoupling decision value of feature j in the i-th welding cycle, which is the core indicator used to quantify the probability of insulation failure.

8. The method for circuit insulation inspection and energization test according to claim 7, characterized in that, The range of the dedicated frequency band is ,in The lower limit frequency; This is the upper limit frequency.

9. The method for circuit insulation inspection and energization test according to claim 7, characterized in that, The method of determining insulation faults through decoupling decision values ​​and fault detection thresholds includes: Extract the historical decision value sequence from the historical normal data and calculate the baseline statistic; Define a drift factor to counteract the progressive characteristic drift caused by electrode aging, namely: ； In the formula, The drift factor; The initial calibration slope; This is the drift sensitivity coefficient; The formula for calculating the fault detection threshold is: ； In the formula, Let be the fault detection threshold for feature j; Confidence factor; The historical decision value of feature j is the average value. The standard deviation of the historical decision values ​​for feature j; when At that time, and when When this occurs, it is determined to be a genuine insulation fault; In the formula, g is the detection cycle number; L is the length of the detection window; This is the preset minimum number of continuous over-limit periods; This is a conditional function, i.e.: 。 10. The method for circuit insulation inspection and energization test according to claim 9, characterized in that, The drift rate deviation calculated based on the electrode aging trend component to trigger electrode aging early warning includes: Calculate the drift rate deviation used to accurately quantify the degree of electrode aging, i.e.: ； In the formula, This refers to the drift rate deviation. The standard error of the drift rate; An electrode aging warning is triggered when the drift rate deviation exceeds a preset critical threshold.

Images