A Relay Arc Parameter Identification Method Based on High Sampling Rate and Multi-Resolution

Abstract

This invention belongs to the field of electrical testing technology, specifically relating to a relay arcing parameter identification method based on high sampling rate and multi-resolution. The method includes using the falling edge of the coil voltage as the coarse alignment time, combining the derivative characteristics of the coil voltage and current for fine alignment, and achieving time delay calibration through inter-channel cross-correlation. A multi-resolution pyramid is constructed from the original sampled data; candidate intervals are generated at the coarse resolution layer using an event intensity function, and then mapped step-by-step to the original layer for precise localization. A relative threshold is calculated based on robust statistics of closed and open platforms. The arcing initiation point is determined using multi-condition continuity constraints of voltage, current, and their derivatives. The arcing end point is determined using a stability index and a bounce segment merging rule. The arcing energy is calculated based on segmented energy integration using state masks and bias correction. This method can accurately identify arcing parameters under high-frequency noise and contact bounce environments, improving measurement accuracy and robustness.

Description

Technical Field

[0001] This invention belongs to the field of electrical testing technology, specifically relating to a method for identifying relay arcing parameters based on high sampling rate and multi-resolution. Background Technology

[0002] Relays are core electromagnetic switching devices in industrial control, automotive electronics, aerospace, and power systems. Their reliability directly determines the safe operation of the entire system, and the arcing characteristics of the contacts are a key indicator for evaluating the lifespan and reliability of relays. When relay contacts open or close, inductive and capacitive loads in the circuit can trigger arcing between the contacts. The high-temperature arc can cause ablation, migration, and oxidation of the contact material, which is the main cause of contact failure. Therefore, accurately measuring parameters such as arcing time and arcing energy is of great significance for optimizing contact material selection, improving contact structure design, and evaluating electrical life.

[0003] Existing methods for testing relay arcing parameters have several technical limitations: First, traditional tests often use the falling edge of the coil voltage as the sole trigger reference, without correcting for multi-channel probe delays. Second, threshold settings often rely on fixed absolute voltage values ​​or simple percentage methods, which are poorly adaptable to baseline drift and plateau voltage fluctuations caused by contact materials, load conditions, and noise interference, easily leading to false triggers or missed detections. Furthermore, to handle massive amounts of data at high sampling rates, algorithms often employ global point-by-point scanning, resulting in low computational efficiency and difficulty meeting online real-time processing requirements. Moreover, under complex bounce and noise glitches, the positioning accuracy of the start and end points is poor. Summary of the Invention

[0004] To address the aforementioned shortcomings of existing technologies, this invention provides a method for identifying relay arcing parameters based on high sampling rate and multi-resolution, comprising: S1. Multi-channel synchronous acquisition of relay coil voltage and contact voltage / current signals, coarse triggering based on the falling edge of coil voltage, defining pre-trigger segment / tail segment, and outputting calibration sampling data after fine alignment and time delay calibration; S2. Based on the calibration sampling data obtained in step S1, construct a multi-resolution pyramid, generate arc candidate intervals at the lowest resolution level, and map them to the high resolution level for refined positioning to obtain a multi-scale candidate interval set. S3. Extract the closed plateau waveform and the open plateau waveform from the pre-trigger segment and the tail segment of the contact voltage channel, respectively, and calculate the adaptive voltage threshold based on the robust statistics of the two. S4. Based on the multi-scale candidate interval set obtained in step S2, and combined with the adaptive voltage threshold, search for the starting position that satisfies the joint conditions of contact voltage, contact current and voltage derivative on the multi-resolution multi-scale pyramid from coarse to fine, and determine the arc starting point that meets the continuity requirement. S5. Starting from the arc initiation point, search backwards and comprehensively determine the arc end point based on the stability index of the contact voltage waveform, the amplitude of the contact current, and the proximity of the average voltage value to the disconnection platform. Then, integrate the rebound events according to the segmentation and merging rules to form the arc event interval. S6. Generate an arc state mask based on the arc event interval, define voltage and current biases, calculate the instantaneous power of bias correction, and obtain the arc energy by integrating only over the effective arc point.

[0005] Further improvements to this technical solution include step S1, which includes: S11. Using the falling edge of the relay coil voltage as the coarse alignment time, synchronously acquire the four channels of contact voltage, coil voltage, contact current and coil current. The sampling accuracy is at the nanosecond level, and define the pre-trigger segment and tail segment as the basis for waveform extraction. S12. Within the local window before and after the coarse alignment time, combine the minimum point of the coil voltage derivative with the characteristic point of the coil current derivative, and achieve sub-sampling level fine alignment through linear interpolation or parabolic fitting to obtain the fine alignment time. S13. Using the pre-trigger segment waveform, perform cross-correlation calculations on the contact voltage channel and contact current channel relative to the coil voltage channel to estimate the sampling offset between channels, and perform translation correction on the time axis to eliminate time base error, thereby obtaining the sampling data after time delay calibration.

[0006] Further improvements to this technical solution include step S2, which includes: S21. Construct a multi-resolution hierarchical pyramid structure based on the sampled data after time delay calibration, including the original layer L0 that maintains the original sampling accuracy, the medium resolution layer L1 that is downsampled using the first downsampling coefficient K1, and the coarse resolution layer L2 that is downsampled using the second downsampling coefficient K2. S22. Define the event intensity function S(t) on the coarse resolution layer L2: ; in, The contact voltage variance is defined with respect to t and a preset window size. This represents the mean of the absolute values ​​of the voltage derivatives within the window. These are the weighting coefficients; S23. Scan the event intensity function S(t) on the coarse resolution layer L2, mark the continuous intervals where the event intensity function S(t) exceeds the preset intensity threshold S_th as arc candidate intervals, and map the arc candidate intervals to the medium resolution layer L1 and the original layer L0 for boundary refinement and precise positioning.

[0007] Further improvements to this technical solution include step S3, which includes: S31. Extract the closed plateau waveform from the pre-trigger segment of the contact voltage channel and calculate the median of the closed plateau voltage. The median of the pre-trigger segment voltage is given, and the median absolute deviation of the closed plateau is calculated. ;in, This represents the voltage value at each sampling point in the contact voltage channel. S32. Extract the disconnection plateau waveform from the tail segment after triggering of the contact voltage channel, and calculate the median of the disconnection plateau voltage. The median of the tail-end voltage is given, and the absolute deviation of the median of the disconnection plateau is calculated. ; S33. Based on robust statistics of closed and open plateaus, calculate the adaptive voltage threshold according to the dual-plateau reference formula. ,in, For platform span coefficient, This is the noise multiplier.

[0008] Further improvements to this technical solution include step S4, which includes: S41. Based on the multi-scale candidate interval set obtained in step S2, the candidate intervals are mapped step by step from the coarse resolution layer L2 to the medium resolution layer L1 and the original layer L0 to obtain a precisely located candidate interval set for starting point search. S42. Within the candidate interval of the original layer L0, calculate the first derivative of the contact voltage. And based on the adaptive voltage threshold of step S3 Set current threshold and voltage derivative threshold As a criterion for judgment; S43. Scan point by point from the starting point of the candidate interval backward, and check the contact voltage at each sampling point. Contact current and voltage change rate A multi-condition joint judgment is made when a preset number of points are continuously satisfied. When the first sampling point that meets the conditions is selected, the arc initiation point is determined. .

[0009] Further improvements to this technical solution include step S5, which includes: S51. Starting from the arc initiation point determined in step S4, set a sliding window backwards and calculate the stability index of the contact voltage within the window: ; in, These are the weighting coefficients; , These represent the maximum and minimum voltage values ​​within the sliding window, respectively. S52. When the sliding window simultaneously satisfies the stability index Contact current And the average window voltage When, mark as a candidate endpoint And verify that the above conditions are consistently met for the minimum stay time. ;in, The preset stability index; Indicates the voltage tolerance of the disconnect platform; S53, If the voltage exceeds the threshold again after the candidate termination point. If the arc initiation conditions are met, it is identified as a secondary arcing event caused by a bounce, and the segmented merging rule is applied to handle events where the interval time is less than the merging threshold. Arc ignition event.

[0010] Further improvements to this technical solution include step S53, which specifically includes: S531. After the candidate end point marked in step S52, continue monitoring backwards for a preset length. If the contact voltage does not exceed the threshold again within the sampling point sequence, then... If the arc initiation condition in step S4 is not met, then the candidate termination point is confirmed as the primary arc termination point. ; S532. If the voltage exceeds the threshold again within the monitoring range. If the arc initiation condition of step S4 is met, it is identified as a secondary arcing event caused by a bounce, and steps S51 and S52 are repeated for the secondary arcing event to determine its corresponding candidate end point. S533. Calculate the time interval gap between the end point of the first arc and the start point of the second arc. If gap is less than the merging threshold... If the two arcs are merged into a single arc event, the end point of the primary arc is updated to the end point of the secondary arc; if If it does not occur, it remains an independent arcing event, and the end point of the main arcing event is still the end point of the first arcing event. S534, Output the starting point of the main arcing event. and end point The starting point is determined in step S4. The endpoint is the final result after processing steps S531 to S533, and the start and end points of each secondary arcing event are output simultaneously.

[0011] Further improvements to this technical solution include step S6, which includes: S61. Generate an arc state mask based on the determined arc start and end points. ,in, This indicates that it is in an arcing state. This indicates that the circuit is in a non-arc state; S62, Define voltage bias Median of closed plateau voltage Define current bias This represents the median of the pre-trigger segment current. S63. Calculate the instantaneous power after bias correction. And the trapezoidal integral method is used only for Energy integration is performed at the sampling points: ; in, This represents the energy integral of the sampling point under arcing conditions. The sampling interval is denoted as . Further improvements to this technical solution include step S6, which further includes: Based on arc state mask Calculate the total arcing time: ; Among them, for all The time intervals corresponding to the sampling points are accumulated; Based on the main arc initiation point determined in step S534 and the end point of the main combustion arc Calculate the main arc time .

[0012] Further improvements to this technical solution include: the method further includes: performing quality scoring on the identification results of steps S1 to S6 and generating a reason code; automatically adjusting the corresponding algorithm parameters according to the reason code category and then triggering re-collection.

[0013] The beneficial effects of this invention are as follows: Step S1 uses subsampling-level fine alignment of coil voltage and current derivative characteristics, combined with pre-trigger segment cross-correlation delay calibration, to control the time base error between multiple channels to the nanosecond level. This completely solves the positioning deviation problem caused by the uncalibrated delay of the traditional single trigger reference, laying the foundation for accurate determination of the arc start and end points.

[0014] Step S3 calculates an adaptive threshold based on the median and absolute deviation of the median between the closed and open platforms, replacing the traditional fixed threshold or mean standard deviation mode. This effectively resists baseline drift caused by differences in contact materials, load fluctuations, and noise spikes, avoiding false triggering and missed detection, and adapting to the testing needs of different relay models.

[0015] Step S2 constructs a three-level multi-resolution pyramid, which quickly generates candidate intervals through a coarse resolution layer and then refines them step by step. Combined with complex window statistical calculations, the processing load of massive high-sampling-rate data is greatly reduced, meeting the needs of online real-time processing. At the same time, the coarse-to-fine search strategy avoids the inefficiency of global point-by-point scanning, improving positioning efficiency while ensuring accuracy.

[0016] Step S4 uses a multi-condition joint judgment based on voltage, current, and voltage derivative, along with continuity constraints, to effectively filter false triggers caused by high-frequency noise spikes. Step S5 quantifies the stability of the waveform through a stability index, and, combined with dwell time verification and bounce segment merging rules, accurately distinguishes between the true end of arcing and contact bounce, supporting independent identification and integration of secondary arcing, thus completely solving the problem of inaccurate start and end point positioning under complex interference.

[0017] Step S6 ensures that integration is performed only on the effective arcing interval by using an arcing state mask, and eliminates the influence of baseline drift by combining voltage and current bias correction. It also uses the trapezoidal integration method to accurately calculate the arcing energy. At the same time, it can output the total arcing energy, the main arcing energy, and the secondary arcing energy separately. The independent calculation of the total arcing time and the main arcing time further improves the precision of parameter measurement and provides more reliable data support for relay life assessment.

[0018] By scoring the quality of the identification results and generating cause codes, targeted parameter self-tuning and resampling closed-loop control are achieved. Compared with the traditional simple resampling strategy, this significantly improves the data consistency and reliability in unattended batch testing, reduces the cost of manual intervention, and adapts to the actual needs of industrial automation testing scenarios. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic flowchart of a method according to an embodiment of the present invention.

[0021] Figure 2 This is a timing diagram for testing the relay release process provided in an embodiment of the present invention.

[0022] Figure 3 This is a schematic diagram of the multi-resolution pyramid and candidate interval generation provided in an embodiment of the present invention.

[0023] Figure 4 This is a flowchart of robust threshold and dual-platform reference calculation provided in an embodiment of the present invention.

[0024] Figure 5 This is a flowchart of composite trigger fine alignment and channel delay calibration provided in an embodiment of the present invention.

[0025] Figure 6 This is a flowchart of the multi-condition continuity determination of the arc initiation point provided in an embodiment of the present invention.

[0026] Figure 7 This is a schematic diagram illustrating the principle of merging stability index and bounce segmentation determination provided in this embodiment of the invention.

[0027] Figure 8 This is a schematic diagram of state mask and bias correction energy calculation provided in an embodiment of the present invention.

[0028] Figure 9 This is a schematic diagram of the cause code-driven parameter self-tuning and re-sampling closed-loop state machine provided in an embodiment of the present invention. Detailed Implementation

[0029] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings of the specific embodiments. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0031] Figure 1 This is a schematic flowchart illustrating a relay arcing parameter identification method based on high sampling rate and multi-resolution provided by the present invention. The order of the steps in this flowchart can be changed, and some steps can be omitted, depending on different requirements.

[0032] like Figure 1 As shown, the method includes: S1. Multi-channel synchronous acquisition of relay coil voltage and contact voltage / current signals, coarse triggering based on the falling edge of coil voltage, defining pre-trigger segment / tail segment, and outputting calibration sampling data after fine alignment and time delay calibration; S2. Based on the calibration sampling data obtained in step S1, construct a multi-resolution pyramid, generate arc candidate intervals at the lowest resolution level, and map them to the high resolution level for refined positioning to obtain a multi-scale candidate interval set. S3. Extract the closed plateau waveform and the open plateau waveform from the pre-trigger segment and the tail segment of the contact voltage channel, respectively, and calculate the adaptive voltage threshold based on the robust statistics of the two. S4. Based on the multi-scale candidate interval set obtained in step S2, and combined with the adaptive voltage threshold, search for the starting position that satisfies the joint conditions of contact voltage, contact current and voltage derivative on the multi-resolution multi-scale pyramid from coarse to fine, and determine the arc starting point that meets the continuity requirement. S5. Starting from the arc initiation point, search backwards and comprehensively determine the arc end point based on the stability index of the contact voltage waveform, the amplitude of the contact current, and the proximity of the average voltage value to the disconnection platform. Then, integrate the rebound events according to the segmentation and merging rules to form the arc event interval. S6. Generate an arc state mask based on the arc event interval, define voltage and current biases, calculate the instantaneous power of bias correction, and obtain the arc energy by integrating only over the effective arc point.

[0033] To facilitate understanding of the present invention, the following description further illustrates the relay arcing parameter identification method based on high sampling rate and multi-resolution, using the principle of the present invention and the process of identifying relay arcing parameters based on high sampling rate and multi-resolution in the embodiments.

[0034] First, step S1 includes: S11. Using the falling edge of the relay coil voltage as the coarse alignment time, synchronously acquire the four channels of contact voltage, coil voltage, contact current and coil current. The sampling accuracy is at the nanosecond level, and define the pre-trigger segment and tail segment as the basis for waveform extraction. S12. Within the local window before and after the coarse alignment time, combine the minimum point of the coil voltage derivative with the characteristic point of the coil current derivative, and achieve sub-sampling level fine alignment through linear interpolation or parabolic fitting to obtain the fine alignment time. S13. Using the pre-trigger segment waveform, perform cross-correlation calculations on the contact voltage channel and contact current channel relative to the coil voltage channel to estimate the sampling offset between channels, and perform translation correction on the time axis to eliminate time base error, thereby obtaining the sampling data after time delay calibration.

[0035] The hardware connections and channel allocation are as follows: The oscilloscope is configured for four-channel synchronous acquisition mode, with the following channel assignments: First channel C1: Connects to the contact voltage detection circuit to measure the voltage across the relay contacts; Second channel C2: Connects to the coil voltage detection circuit to measure the voltage across the relay coil and also serves as a trigger source; The third channel C3: connected to the contact current detection circuit, which measures the contact current through a sampling resistor or current transformer; The fourth channel, C4, is connected to the coil current detection circuit to measure the coil excitation current.

[0036] The trigger configuration uses the falling edge trigger of the C2 channel, the trigger level is set to 50% of the coil's rated voltage, and the pre-trigger ratio is set to 10% to ensure that the baseline waveform before triggering can be acquired for time delay calibration and threshold calculation.

[0037] Step S1, Composite Trigger Fine Alignment and Channel Delay Calibration: The falling edge of the relay coil voltage is used as the coarse alignment time. ,like Figure 2 As shown, four channels—contact voltage, coil voltage, contact current, and coil current—are synchronously acquired with a sampling accuracy at the nanosecond level. Specifically, contact voltage is assigned to the first acquisition channel C1, coil voltage to the second acquisition channel C2, contact current to the third acquisition channel C3, and coil current to the fourth acquisition channel C4. Front and back partial windows Internal, combined coil voltage derivative The minimum point and the derivative of the coil current The feature points are used to achieve subsampling-level fine alignment through linear interpolation, and the fine alignment time is obtained. By using the waveforms of each channel in the pre-triggered segment to perform cross-correlation calculations or feature point alignment on C1 and C3 relative to C2, the sampling offset of each channel can be estimated. and It also corrects the time axis of the corresponding channel to eliminate time base errors caused by inter-channel propagation delay and probe delay.

[0038] In step S1, the sampling accuracy is 80 nanoseconds per point, and the number of sampling points in a single acquisition is not less than 1 million points; the value range of the local window ΔT is 10μs to 100μs.

[0039] like Figure 5 As shown, the process for composite trigger fine alignment and channel delay calibration is as follows: Coarse alignment timing determination: The falling edge trigger time of channel C2 (coil voltage) is used as the coarse alignment timing. The oscilloscope's hardware triggering mechanism guarantees repeatability at that moment, but due to trigger circuit delays and the discreteness of digital sampling, There is uncertainty for a maximum sampling period.

[0040] Fine alignment window settings: In Set up precise alignment windows before and after, with the window range being [missing information]. In this embodiment A 50μs interval is used, corresponding to 625 sampling points.

[0041] Derivative feature extraction: Calculate the first derivative of the voltage of the C2 channel coil within the finely aligned window. and the first derivative of the current in the C4 channel coil The derivative is calculated using the central difference method: ; Fine alignment time location: Within the fine alignment window, search The minimum point (corresponding to the moment when the voltage drops the fastest) and Feature points (e.g., zero-crossing points or extreme points). If the two positions are close (difference less than 5 sampling points), the average position of the two is taken as the candidate point for fine alignment; otherwise, the average position is preferred. Minimal point.

[0042] Subsampling interpolation: Parabolic fitting or linear interpolation is performed on the derivative sequence near the fine alignment candidate points to obtain the fine alignment time with subsampling accuracy. The interpolation accuracy can reach 1 / 10 of the sampling interval, i.e., on the order of 8 ns.

[0043] Channel delay estimation: using pre-triggered segments ( Channel delay estimation is performed on the waveform of (C1, C3, and C2). Cross-correlation is then performed on channels C1 and C3 relative to channel C2. ; ; The offset corresponding to the cross-correlation peak is the sampling offset between channels. and .

[0044] Time axis calibration: Based on the estimated sampling offset, the time axes of channels C1 and C3 are shifted and corrected. ; ; After calibration, the time base of each channel is aligned to channel C2, eliminating the time base error caused by probe delay and signal link delay.

[0045] The technical advantage of channel delay calibration lies in the fact that probes, cables, and signal conditioning circuits in different channels have different propagation delays, typically ranging from tens to hundreds of nanoseconds. Without delay calibration, the positioning of the arc initiation and termination points on different channels will be inaccurate, affecting the consistency of the joint determination between the two channels. By estimating and calibrating the delay through cross-correlation, the inter-channel time base error can be controlled within one sampling period.

[0046] Secondly, step S2 includes: S21. Construct a multi-resolution hierarchical pyramid structure based on the sampled data after time delay calibration, including the original layer L0 that maintains the original sampling accuracy, the medium resolution layer L1 that is downsampled using the first downsampling coefficient K1, and the coarse resolution layer L2 that is downsampled using the second downsampling coefficient K2. S22. Define the event intensity function S(t) on the coarse resolution layer L2: ; in, The contact voltage variance is defined with respect to t and a preset window size. This represents the mean of the absolute values ​​of the voltage derivatives within the window. These are the weighting coefficients; S23. Scan the event intensity function S(t) on the coarse resolution layer L2, mark the continuous intervals where the event intensity function S(t) exceeds the preset intensity threshold S_th as arc candidate intervals, and map the arc candidate intervals to the medium resolution layer L1 and the original layer L0 for boundary refinement and precise positioning.

[0047] Step S2, Multi-scale Candidate Interval Generation and Complexity-Controlled Pyramid Construction: The time-delay calibrated sampled data obtained in Step S1 is decomposed into a multi-resolution structure to construct a three-level pyramid structure. The original layer L0 maintains the original sampling accuracy for precise positioning and energy calculation. The medium-resolution layer L1 uses a first downsampling coefficient K1 to downsample the original data, and the coarse-resolution layer L2 uses a second downsampling coefficient K2. The first and second downsampling coefficients K1 and K2 can be adaptively selected within a preset range based on noise estimation or the desired arcing timescale. The value range of K1 is 4 to 20, and the value range of K2 is 16 to 200. An event intensity function S(t) is defined. The candidate interval set is generated by scanning S(t) on the coarse-resolution layer L2. The event intensity function S(t) is calculated as a weighted sum of the window variance and the window derivative energy. ,in Indicates window variance. Represents the window mean. The weight coefficient ranges from 0.1 to 10; the candidate intervals of the L2 layer are mapped to the L1 layer for refinement and screening, and then mapped to the L0 layer for precise positioning; the downsampling adopts the window mean method, and the window mean and variance are updated in O(1) complexity through prefix sum or sliding update, which is suitable for the real-time processing requirements of millions of data points.

[0048] In step S2, the first downsampling coefficient K1 is 10, corresponding to an equivalent sampling precision of 800 nanoseconds per point for the medium resolution layer L1; the second downsampling coefficient K2 is 100, corresponding to an equivalent sampling precision of 8 microseconds per point for the coarse resolution layer L2; and the weighting coefficient λ is 1.0.

[0049] like Figure 3 As shown, the process of generating the multi-resolution pyramid and candidate intervals is as follows: The original layer L0: Keeps the original sampling data unchanged, with a sampling accuracy of 80 ns / point and a data point count of N0 (approximately 1.25 million points). The original layer is used for final precise positioning and energy calculation.

[0050] Medium resolution layer L1: The original data is downsampled by a factor of K1. K1 can be adaptively selected within the range of 4 to 20; in this embodiment, K1=10. A window averaging method is used to average 10 consecutive sampling points to obtain one L1 layer data point. The equivalent sampling precision of the L1 layer is 800 ns / point, and the number of data points is N1=N0 / 10 (approximately 125,000 points).

[0051] Coarse resolution layer L2: The original data is downsampled by a factor of K2. K2 can be adaptively selected in the range of 16 to 200; in this embodiment, K2=100. One L2 layer data point is obtained by averaging 100 consecutive sampling points. The equivalent sampling precision of the L2 layer is 8 μs / point, and the number of data points is N2=N0 / 100 (approximately 12,500 points).

[0052] Adaptive selection of downsampling coefficients: The selection of K1 and K2 can be adaptively determined according to the following strategy: Based on noise estimation: If the noise in the pre-trigger segment is large (MAD>MAD_th), appropriately increase K1 and K2 to enhance the low-pass filtering effect; Based on the expected arcing timescale: if the typical arcing time of the relay under test is short (<1ms), K2 should be appropriately reduced to retain more time resolution.

[0053] Event intensity function definition: In the L2 layer, an event intensity function S(t) is defined to identify potential arcing regions. ; in: For Centered on, window size is The contact voltage variance; This represents the mean of the absolute values ​​of the voltage derivatives within the window; As the weighting coefficient, this embodiment takes... The value ranges from 0.1 to 10.

[0054] Fast computation implementation: To achieve real-time processing of millions of data points, the window mean and variance are calculated using prefix sums or sliding update methods. Prefix sum: ; Window mean: ; Window variance: ; By pre-compiling with prefix sums, the computational complexity of the mean and variance for each window is O(1), and the overall complexity is O(N).

[0055] Candidate interval generation: Scan the event intensity function S(t) at layer L2, set an intensity threshold S_th, and mark continuous intervals where S(t) > S_th as candidate intervals. Merge adjacent candidate intervals to avoid excessive fragmentation.

[0056] Stepwise mapping refinement: Candidate intervals from layer L2 are mapped to layer L1 according to the index ratio. At layer L1, the candidate intervals are refined, validated, and their boundaries fine-tuned. Then, the fine-tuned intervals from layer L1 are mapped to layer L0 for final precise localization. The mapping relationship is as follows: L0 index = L1 index × K1 = L2 index × K2.

[0057] In addition, step S3 includes: S31. Extract the closed plateau waveform from the pre-trigger segment of the contact voltage channel and calculate the median of the closed plateau voltage. The median of the pre-trigger segment voltage is given, and the median absolute deviation of the closed plateau is calculated. ;in, This represents the voltage value at each sampling point in the contact voltage channel. S32. Extract the disconnection plateau waveform from the tail segment after triggering of the contact voltage channel, and calculate the median of the disconnection plateau voltage. The median of the tail-end voltage is given, and the absolute deviation of the median of the disconnection plateau is calculated. ; S33. Based on robust statistics of closed and open plateaus, calculate the adaptive voltage threshold according to the dual-plateau reference formula. ,in, For platform span coefficient, This is the noise multiplier.

[0058] Step S3, Robust Statistics and Adaptive Threshold Calculation with Dual-Platform Reference: Extract the closed-plateau waveform from the pre-trigger segment of contact voltage channel C1, and calculate the closed-plateau voltage. The median of the pre-trigger segment voltage is used; the disconnection plateau waveform is extracted from the tail segment after triggering, and the disconnection plateau voltage is calculated. The median of the tail-end voltage is used; the median absolute deviation (MAD) is used as the noise scale estimate, and the calculation formula is as follows: ; Calculate the adaptive voltage threshold based on the dual-platform reference formula ,in This is the platform span coefficient, with a value ranging from 0.05 to 0.3. The noise multiplier coefficient has a value range of 2 to 6; the median and median absolute deviation are more robust than the mean and standard deviation, and can effectively suppress the interference of spikes, bounces and outliers on threshold calculation.

[0059] In step S3, the platform span coefficient α is 0.1, the noise multiplier coefficient k is 3, the number of sampling points in the pre-trigger segment is 5000, and the corresponding time length is 400μs.

[0060] like Figure 4 As shown, the robust threshold calculation process is as follows: Closed-plate extraction: Extract pre-trigger segment data from contact voltage channel C1. The pre-trigger segment is defined as the fine alignment time. Previous Each sampling point, in this embodiment The corresponding time length is 400μs. The pre-trigger segment represents the baseline voltage level when the contacts are closed.

[0061] Robust statistics of closed platforms: Robust statistics of closed platforms are calculated using the median and median absolute deviation. Closed plateau voltage: ; Median absolute deviation: ; Compared to the mean, the median is more resistant to outliers; compared to the standard deviation, the MAD is not affected by extreme values ​​and can accurately reflect the typical dispersion of the data.

[0062] Platform disconnection extraction: Extract the tail segment data from the contact voltage channel C1. The tail segment is defined as the Nt sampling points at the end of the acquisition window; in this embodiment, Nt = 5000. The tail segment represents the steady-state voltage level after the contact is completely disconnected.

[0063] Robust statistics for disconnecting from the platform: Disconnect platform voltage: ; Median absolute deviation: ; Dual-platform reference threshold calculation: Calculate the adaptive voltage threshold according to the dual-platform reference formula: ; in: The platform span coefficient represents the position of the threshold relative to the closing-opening voltage span. In this embodiment, it is taken as... The value ranges from 0.05 to 0.3; k is the noise multiplier coefficient, which is taken as in this embodiment. The value ranges from 2 to 6.

[0064] Threshold validity verification: The calculated threshold is validated for reasonableness. like (e.g., 5V) indicates that the difference between the closed and open platforms is too small, which may indicate a wiring or data acquisition abnormality, and a quality warning will be issued. like (e.g., 200V), using the preset default threshold and marking it as abnormal.

[0065] Current threshold calculation: Similarly, current threshold Robust statistics based on the pre-trigger segment current can be used to calculate: ; ; ; In this embodiment, the following is taken If the calculated result is less than 0.05A, then 0.05A is taken as the minimum current threshold.

[0066] The advantage of robust thresholding lies in the fact that the traditional mean plus standard deviation method is easily affected by spikes and outliers—a single large outlier can significantly increase the standard deviation, leading to threshold distortion. By using the median and MAD, the threshold calculation remains stable even with a small number of outliers, improving consistency across batch measurements.

[0067] Next, step S4 includes: S41. Based on the multi-scale candidate interval set obtained in step S2, the candidate intervals are mapped step by step from the coarse resolution layer L2 to the medium resolution layer L1 and the original layer L0 to obtain a precisely located candidate interval set for starting point search. S42. Within the candidate interval of the original layer L0, calculate the first derivative of the contact voltage. And based on the adaptive voltage threshold of step S3 Set current threshold and voltage derivative threshold As a criterion for judgment; S43. Scan point by point from the starting point of the candidate interval backward, and check the contact voltage at each sampling point. Contact current and voltage change rate A multi-condition joint judgment is made when a preset number of points are continuously satisfied. When the first sampling point that meets the conditions is selected, the arc initiation point is determined. .

[0068] Step S4, Joint Determination of Arc Initiation Point Based on Continuity and Derivative Constraints: The arc initiation point is located on the multi-resolution pyramid using a coarse-to-fine search strategy. First, candidate intervals are identified on the coarse-resolution layer L2 based on the event intensity function S(t). These candidate intervals are then mapped to the medium-resolution layer L1 for refinement, and finally, precise location is achieved on the original layer L0. During location on the original layer L0, a multi-condition joint determination is used, considering contact voltage, contact current, and voltage derivative, to determine the arc initiation point. The following conditions must be met simultaneously: contact voltage Contact current Voltage change rate Furthermore, the above conditions must be met consecutively. Each sampling point or duration ;in This is the current threshold, ranging from 0.01 amperes to 0.5 amperes; The voltage derivative threshold value ranges from 1V / μs to 100V / μs. The consecutive point threshold ranges from 3 to 20 points, corresponding to the duration. The range is from 0.24 μs to 1.6 μs; the continuity constraint can effectively filter out false triggers caused by transient noise spikes.

[0069] In step S4, the current threshold Ith is 0.05 amperes, the voltage derivative threshold Dv_th is 10 V / μs, and the continuous point threshold N_start is 5 points.

[0070] like Figure 6 As shown, the multi-condition continuity determination process for the arc initiation point is as follows: Candidate interval acquisition: Obtain the set of candidate intervals mapped to the L0 layer from the multi-scale candidate interval generation module.

[0071] Voltage derivative calculation: Calculate the first derivative of the contact voltage within the candidate interval: ; Where Δt = 80 ns is the sampling interval.

[0072] Multi-condition joint judgment: Scan point by point from the starting point of the candidate interval, and check the following three conditions for each sampling point: Condition 1 (Voltage Condition): ; Condition 2 (Current Condition): ; Condition 3 (Derivative Condition): .

[0073] In this embodiment, , .

[0074] Continuity constraint verification: Counting begins when a sampling point simultaneously meets all three of the above conditions; if continuous... If all sampling points meet the three conditions, then the first sampling point that meets the conditions is determined as the arc initiation point. .

[0075] In this embodiment Corresponding duration .

[0076] Confidence score (optional): A confidence score function can be defined to quantitatively evaluate the arc initiation determination. ; in This is the weighting coefficient. When the score continuously exceeds the scoring threshold... At that time, it is determined to be the start of an arc.

[0077] The technical advantages of multi-condition continuity constraints are: The voltage conditions ensure that the contact voltage has risen significantly, eliminating baseline fluctuations; The current condition ensures that current still flows between the contacts (maintaining the arc) and eliminates interference from induced voltage. The derivative condition ensures that the voltage is changing rapidly, eliminating the possibility of misjudging steady-state high voltage. Continuity constraints ensure that conditions are met stably rather than transient spikes, effectively filtering high-frequency noise.

[0078] In addition, step S5 includes: S51. Starting from the arc initiation point determined in step S4, set a sliding window backwards and calculate the stability index of the contact voltage within the window: ; in, These are the weighting coefficients; , These represent the maximum and minimum voltage values ​​within the sliding window, respectively. S52. When the sliding window simultaneously satisfies the stability index Contact current And the average window voltage When, mark as a candidate endpoint And verify that the above conditions are consistently met for the minimum stay time. ;in, The preset stability index; This indicates the voltage tolerance of the disconnection platform, which is the allowable deviation range for determining whether the average voltage window value is close to the voltage of the disconnection platform. S53, If the voltage exceeds the threshold again after the candidate termination point. If the arc initiation conditions are met, it is identified as a secondary arcing event caused by a bounce, and the segmented merging rule is applied to handle events where the interval time is less than the merging threshold. Arc ignition event.

[0079] Furthermore, step S53 specifically includes: S531. After the candidate end point marked in step S52, continue monitoring backwards for a preset length. If the contact voltage does not exceed the threshold again within the sampling point sequence, then... If the arc initiation condition in step S4 is not met, then the candidate termination point is confirmed as the primary arc termination point. ; S532. If the voltage exceeds the threshold again within the monitoring range. If the arc initiation condition of step S4 is met, it is identified as a secondary arcing event caused by a bounce, and steps S51 and S52 are repeated for the secondary arcing event to determine its corresponding candidate end point. S533. Calculate the time interval gap between the end point of the first arc and the start point of the second arc. If gap is less than the merging threshold... If the two arcs are merged into a single arc event, the end point of the primary arc is updated to the end point of the secondary arc; if If it does not occur, it remains an independent arcing event, and the end point of the main arcing event is still the end point of the first arcing event. S534, Output the starting point of the main arcing event. and end point The starting point is determined in step S4. The endpoint is the final result after processing steps S531 to S533, and the start and end points of each secondary arcing event are output simultaneously.

[0080] Step S5, Determination of the arc termination point by merging the stability index and the rebound segment: From the arc start point Search backwards for the end point of the arc. The stability index SI is defined to quantify the stability of a voltage waveform, and its calculation formula is as follows: ,in For window variance, The mean of the window. and These are the maximum and minimum values ​​within the window, respectively. , These are normalized weighting coefficients, all ranging from 0 to 1. A point is considered a candidate endpoint when it simultaneously meets the following conditions: stability index. ,in The stability threshold value ranges from 0.01 to 0.5; contact current. Window mean Approaching the disconnection platform voltage, satisfying ,in For platform tolerance, the value ranges from 0.5V to 5V; the above conditions must be met continuously for no less than the minimum dwell time. The value ranges from 100μs to 1000μs; for cases with bounce, a segmented merging rule is adopted: if the interval between two arcing events is less than the merging threshold... If the arcing time is within the specified range, then merge them into a single arcing event; otherwise, record them as independent secondary arcing events and output the arcing time of each segment separately. and arc energy ; The value ranges from 50μs to 500μs.

[0081] In step S5, the stability index weighting coefficients a is 0.4, b is 0.3, and c is 0.3; the stability threshold SI_th is 0.1; and the platform tolerance... The minimum residence time is 2V. The value is 400 μs, corresponding to a sliding window size W of 5000 points; the merging threshold τ_merge is 200 μs.

[0082] In step S5, during the determination process, a moving average filter can also be used to preprocess the contact voltage. The moving average window size can be selected as 5 points.

[0083] like Figure 7 As shown, the process for determining the stability index of the arc termination point and the bounce segment is as follows: Stability index calculation: from the arc initiation point Set a sliding window to the left, with a window size of [size missing]. Each sampling point (in this embodiment) (corresponding to 400 μs). Calculate the stability index SI for each window position: ; in: This represents the voltage variance within the window, reflecting the degree of voltage fluctuation. This represents the average absolute value of the voltage derivative within the window, reflecting the rate of voltage change. This represents the voltage range within the window, reflecting the voltage fluctuation range. For the normalized weighting coefficients, this embodiment takes... .

[0084] Candidate End Point Detection: A sliding window is marked as a candidate end point when it meets the following conditions: Stability Index: ; Current condition: ; Platform proximity: .

[0085] Validation of dwell time: The above conditions for the candidate endpoint must be continuously met to ensure the minimum dwell time. (This embodiment) If in If the conditions are no longer met within the time limit, the candidate endpoint is cancelled and the search continues.

[0086] Jumpback detection and segmentation: After confirming the candidate end point, continue monitoring forward for a distance (e.g., 1000 points). If the voltage exceeds the threshold again... If the arc initiation condition is met, it is identified as a secondary arcing event caused by a bounce. The above steps are repeated for secondary arcing events until a truly stable termination point is found.

[0087] Segment merging rule: For the identified multi-segment arcing events, the segment merging rule is applied: If the interval between the two arcing stages (This embodiment) If they are combined into a single arcing event, the total arcing time is from the start of the first segment to the end of the last segment. like If the arc is isolated, the start and end times, arc time, and arc energy of each segment are recorded separately.

[0088] Output: Output the start and end points of the main arcing event. and the start and end points of each secondary arcing event. (If it exists).

[0089] The technical advantages of the stability index are: compared to the traditional "window variance + residence time + hysteresis" triad, the stability index integrates multiple stability measures into a single indicator, with adjustable weights, making it more flexible; platform proximity judgment is based on... With a tolerance band instead of a fixed threshold (e.g., 0.5V), it can adapt to relay testing at different voltage levels.

[0090] Finally, step S6 includes: S61. Generate an arc state mask based on the determined arc start and end points. ,in, This indicates that it is in an arcing state. This indicates that the circuit is in a non-arc state; S62, Define voltage bias Median of closed plateau voltage Define current bias This is the median of the pre-trigger segment current, used to eliminate the influence of baseline and DC offset on power calculation; S63. Calculate the instantaneous power after bias correction. And the trapezoidal integral method is used only for Energy integration is performed at the sampling points: ; in, This represents the energy integral of the sampling point under arcing conditions. The sampling interval is denoted as . Furthermore, step S6 also includes: Based on arc state mask Calculate the total arcing time: ; Among them, for all The time intervals corresponding to the sampling points are accumulated; Based on the main arc initiation point determined in step S534 and the end point of the main combustion arc Calculate the main arc time .

[0091] Step S6, Segmented Calculation of Arcing Energy Based on State Mask and Offset Correction: Based on the arcing start and end points determined in steps S4 and S5, generate the arcing state mask. ,in This indicates that it is in an arcing state. Indicates a non-arc state; defines voltage bias. Closed plateau voltage Or the median of the pre-trigger segment voltage, defining the current bias. The median of the pre-triggered segment current or the baseline of the closed-state current is used; the instantaneous power after bias correction is calculated. Only for Energy is integrated at the sampling points, and the arc energy is calculated using the trapezoidal integration method: ,in The sampling interval is specified; the total arc energy is output. Main combustion arc energy And the energy of each secondary arc or rebound arc. (If present); Arcing time for The number of sampling points multiplied by the sampling interval.

[0092] like Figure 8As shown, the calculation process for arc energy using state mask and bias correction is as follows: Arc state mask generation: Based on the arc start and end point determination results, generate an arc state mask array arc_mask(t) of the same length as the original sampling sequence. like , ;otherwise, .

[0093] If there are multiple arcing events, the mask value in all arcing intervals is 1.

[0094] Bias value determination: Voltage bias: (Median of closed-plate voltage); Current bias: (Median of pre-trigger segment current); The bias value represents the baseline level under non-arc conditions. Subtracting the bias value can eliminate the effects of DC offset and baseline drift.

[0095] Offset-corrected power calculation: Calculate the instantaneous power after offset correction for each sampling point: ; Mask integral: only for Energy integration is performed at the sampling points using the trapezoidal integration method: ; in The sampling interval is denoted as .

[0096] Segmented energy output: Total arc energy: The sum of integrals between; Main arc energy: The integral value of the main combustion arc interval; Secondary arc energy: The integral value of the second arc interval (if it exists).

[0097] Arc time calculation: Total arc burning time: (The time corresponding to all sampling points with a mask of 1); Main arc duration: .

[0098] The technical benefits of state masking and offset correction are: The masking mechanism ensures that integration is performed only on the true arcing region, avoiding the miscalculation of signals from non-arcing regions into the energy calculation. Bias correction eliminates baseline offsets introduced by factors such as contact closure resistance and sampling circuit bias, improving the accuracy of energy calculation. Segmented output supports independent analysis of secondary arcing caused by bounce, facilitating in-depth study of contact mechanical characteristics.

[0099] In addition, the method also includes: scoring the recognition results of steps S1 to S6 and generating a reason code; automatically adjusting the corresponding algorithm parameters according to the reason code category and then triggering re-collection.

[0100] Step S7, Parameter self-tuning and quality closed-loop control driven by reason code: The test is scored, and the total quality score Q and reason code are output. The reason code is an enumeration type, including at least the following categories: RC_TRIGGER_FAIL (triggered alignment failure), RC_DELAY_ABNORMAL (abnormal channel delay), RC_NOISE_HIGH (excessive noise), RC_PLATFORM_UNCERTAIN (platform uncertainty), RC_BOUNCE_EXCESS (excessive bounce), RC_WINDOW_INSUFFICIENT (insufficient sampling window). The corresponding parameters are automatically adjusted according to the reason code, and then re-acquisition is triggered: when the reason code is RC_NOISE_... When the HIGH level is reached, increase the downsampling coefficient K2 or increase the noise multiplier k. When the cause code is RC_TRIGGER_FAIL, adjust the trigger level or increase the pre-trigger ratio. When the cause code is RC_PLATFORM_UNCERTAIN, increase the minimum dwell time τmin or adjust the stability threshold SI_th or platform tolerance εopen. When the cause code is RC_WINDOW_INSUFFICIENT, increase the acquisition duration or storage depth. If the total score Q is lower than the preset quality threshold Qth, automatically resample after parameter self-tuning according to the cause code. If multiple resamples still fail, mark the sample as abnormal and trigger an alarm. The quality threshold Qth ranges from 60% to 80% of the full score.

[0101] In step S7, the full score is 100 points, and the quality threshold Qth is 70 points.

[0102] like Figure 9 As shown, the cause-code driven parameter self-tuning and resampling closed-loop mechanism is implemented using a state machine, including the following states and transitions: State S0 - Idle state: Waiting for test start command.

[0103] State S1 - Acquisition State: Execute waveform acquisition, and switch to scoring state after acquisition is completed.

[0104] State S2 - Scoring and Diagnostic State: The acquired waveform data is scored in multiple dimensions, and a cause code is generated. The mapping between scoring items and cause codes is shown in Table 1. Table 1:

[0105] Reason code definition: enum ReasonCode { RC_OK = 0, / / No exceptions RC_TRIGGER_FAIL = 1, / / Alignment failed to trigger. RC_DELAY_ABNORMAL = 2, / / Channel delay anomaly RC_NOISE_HIGH = 3, / / Excessive noise RC_PLATFORM_UNCERTAIN = 4, / / Platform uncertain RC_BOUNCE_EXCESS = 5, / / More bounces RC_WINDOW_INSUFFICIENT = 6 / / Insufficient sampling window } State S3 - Decision State: Compare the total score Q with the quality threshold Qth = 70 points. If... and If the result is deemed satisfactory, the system proceeds to the completion state; otherwise, it proceeds to the parameter adjustment state.

[0106] Status S4 - Parameter Adjustment Status: Query the parameter adjustment strategy table based on the reason code, as shown in Table 2, and automatically adjust the corresponding parameters. Table 2: Reason code Adjust strategy RC_TRIGGER_FAIL Reduce the trigger level by 10%, or increase the pre-trigger ratio to 15%. RC_DELAY_ABNORMAL Expand the cross-correlation search range, or use feature point alignment instead of cross-correlation. RC_NOISE_HIGH Increase K2 to 150, or increase the noise multiplier k to 4. RC_PLATFORM_UNCERTAIN Increase τmin to 600 μs, or adjust SI_th to 0.15, or increase εopen to 3V. RC_BOUNCE_EXCESS Increase τ_merge to 300μs to incorporate more bounces. RC_WINDOW_INSUFFICIENT Increase the acquisition duration to 150ms, or adjust the pre-trigger ratio. Status S5 - Resampling Status: The resampling counter increments by 1, and checks if the maximum number of resampling attempts has been exceeded (default is 3). If not exceeded, return to the acquisition status and re-acquire; if exceeded, mark as an abnormal sample and enter abnormal status.

[0107] Status S6 - Completed: Save test results, update statistics, and return to idle status to wait for the next test.

[0108] Status S7 - Abnormal Status: Generates an abnormal alarm, records the complete cause code history and parameter adjustment records, and waits for manual intervention or skips the current sample.

[0109] The advantage of cause-code driven parameter self-tuning lies in its ability to identify specific failure causes and implement targeted parameter adjustments compared to a simple "low-score resampling" strategy. For example, it can increase the filter intensity when noise is excessive or extend the dwell time when the platform is uncertain. This targeted adjustment has higher convergence efficiency, enabling the acquisition of effective results with fewer resampling attempts, thus improving the efficiency and reliability of unattended batch testing.

[0110] Based on the above embodiments, the complete testing process of the present invention is as follows: Step 1: System initialization, configure oscilloscope sampling parameters (sampling rate 12.5MHz, storage depth 1.25 million points, C2 falling edge trigger, pre-trigger ratio 10%).

[0111] Step 2: Connect the relay under test correctly: connect the contact voltage to C1, the coil voltage to C2, the contact current to C3, and the coil current to C4.

[0112] Step 3: Energize the relay coil to make the relay click and the contacts close.

[0113] Step 4: Start the oscilloscope and wait for triggering, then disconnect the coil power supply and trigger the oscilloscope to acquire data.

[0114] Step 5: Read the four-channel waveform data.

[0115] Step 6: Perform compound trigger fine alignment: Using the falling edge of C2 as the coarse alignment time, perform subsampling-level fine alignment within the local window by combining the dVcoil / dt and dIcoil / dt features.

[0116] Step 7: Perform channel delay calibration: Use the pre-trigger segment to estimate the cross-correlation between C1 and C3 and C2, and calibrate the time axis of each channel.

[0117] Step 8: Construct a multi-resolution pyramid (L0 / L1 / L2), and generate candidate intervals by scanning the event intensity function S(t) in the L2 layer.

[0118] Step 9: Extract Vclose and Vopen from the pre-triggered segment and the tail segment respectively, and calculate the robust threshold using the median and MAD. .

[0119] Step 10: Locate the arc initiation point from coarse to fine within the candidate interval, using voltage + current + derivative multi-condition continuity constraints for determination.

[0120] Step 11: Search for the end point of the arcing from the starting point backward, use the stability index SI to determine it, and apply the jump segment merging rule.

[0121] Step 12: Generate the arc state mask arc_mask, calculate and integrate the power after bias correction, and output the total energy. Main combustion arc energy and secondary arc energy .

[0122] Step 13: Perform quality scoring and generate a reason code (reason_code); if the result is unqualified, adjust the parameters according to the reason code and automatically resample; if the result is qualified, save the result.

[0123] Step 14: Repeat steps 3-13 to complete the batch test.

[0124] Although the present invention has been described in detail with reference to the accompanying drawings and preferred embodiments, the present invention is not limited thereto. Various equivalent modifications or substitutions can be made to the embodiments of the present invention by those skilled in the art without departing from the spirit and essence of the invention, and such modifications or substitutions should all be within the scope of the present invention. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should also be covered within the protection scope of the present invention.

Claims

1. A method for identifying relay arcing parameters based on high sampling rate and multi-resolution, characterized in that, include: S1. Multi-channel synchronous acquisition of relay coil voltage and contact voltage / current signals, coarse triggering based on the falling edge of coil voltage, defining pre-trigger segment / tail segment, and outputting calibration sampling data after fine alignment and time delay calibration; S2. Based on the calibration sampling data obtained in step S1, construct a multi-resolution pyramid, generate arc candidate intervals at the lowest resolution level, and map them to the high resolution level for refined positioning to obtain a multi-scale candidate interval set. S3. Extract the closed plateau waveform and the open plateau waveform from the pre-trigger segment and the tail segment of the contact voltage channel, respectively, and calculate the adaptive voltage threshold based on the robust statistics of the two. S4. Based on the multi-scale candidate interval set obtained in step S2, and combined with the adaptive voltage threshold, search for the starting position that satisfies the joint conditions of contact voltage, contact current and voltage derivative on the multi-resolution multi-scale pyramid from coarse to fine, and determine the arc starting point that meets the continuity requirement. S5. Starting from the arc initiation point, search backwards and comprehensively determine the arc end point based on the stability index of the contact voltage waveform, the amplitude of the contact current, and the proximity of the average voltage value to the disconnection platform. Then, integrate the rebound events according to the segmentation and merging rules to form the arc event interval. S6. Generate an arc state mask based on the arc event interval, define voltage and current biases, calculate the instantaneous power of bias correction, and obtain the arc energy by integrating only over the effective arc point.

2. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 1, characterized in that, Step S1 includes: S11. Using the falling edge of the relay coil voltage as the coarse alignment time, synchronously acquire the four channels of contact voltage, coil voltage, contact current and coil current. The sampling accuracy is at the nanosecond level, and define the pre-trigger segment and tail segment as the basis for waveform extraction. S12. Within the local window before and after the coarse alignment time, combine the minimum point of the coil voltage derivative with the characteristic point of the coil current derivative, and achieve sub-sampling level fine alignment through linear interpolation or parabolic fitting to obtain the fine alignment time. S13. Using the pre-trigger segment waveform, perform cross-correlation calculations on the contact voltage channel and contact current channel relative to the coil voltage channel to estimate the sampling offset between channels, and perform translation correction on the time axis to eliminate time base error, thereby obtaining the sampling data after time delay calibration.

3. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 1, characterized in that, Step S2 includes: S21. Construct a multi-resolution hierarchical pyramid structure based on the sampled data after time delay calibration, including the original layer L0 that maintains the original sampling accuracy, the medium resolution layer L1 that is downsampled using the first downsampling coefficient K1, and the coarse resolution layer L2 that is downsampled using the second downsampling coefficient K2. S22. Define the event intensity function S(t) on the coarse resolution layer L2: ； in, The contact voltage variance is defined with respect to t and a preset window size. This represents the mean of the absolute values ​​of the voltage derivatives within the window; These are the weighting coefficients; S23. Scan the event intensity function S(t) on the coarse resolution layer L2, mark the continuous intervals where the event intensity function S(t) exceeds the preset intensity threshold S_th as arc candidate intervals, and map the arc candidate intervals to the medium resolution layer L1 and the original layer L0 for boundary refinement and precise positioning.

4. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 3, characterized in that, Step S3 includes: S31. Extract the closed plateau waveform from the pre-trigger segment of the contact voltage channel and calculate the median of the closed plateau voltage. The median of the pre-trigger segment voltage is given, and the median absolute deviation of the closed plateau is calculated. ;in, This represents the voltage value at each sampling point in the contact voltage channel. S32. Extract the disconnection plateau waveform from the tail segment after triggering of the contact voltage channel, and calculate the median of the disconnection plateau voltage. The median of the tail-end voltage is given, and the absolute deviation of the median of the disconnection plateau is calculated. ; S33. Based on robust statistics of closed and open plateaus, calculate the adaptive voltage threshold according to the dual-plateau reference formula. ,in, For platform span coefficient, This is the noise multiplier.

5. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 4, characterized in that, Step S4 includes: S41. Based on the multi-scale candidate interval set obtained in step S2, the candidate intervals are mapped step by step from the coarse resolution layer L2 to the medium resolution layer L1 and the original layer L0 to obtain a precisely located candidate interval set for starting point search. S42. Within the candidate interval of the original layer L0, calculate the first derivative of the contact voltage. And based on the adaptive voltage threshold of step S3 Set current threshold and voltage derivative threshold As a criterion for judgment; S43. Scan point by point from the starting point of the candidate interval backward, and check the contact voltage at each sampling point. Contact current and voltage change rate A multi-condition joint judgment is made when a preset number of points are continuously satisfied. When the first sampling point that meets the conditions is selected, the arc initiation point is determined. .

6. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 5, characterized in that, Step S5 includes: S51. Starting from the arc initiation point determined in step S4, set a sliding window backwards and calculate the stability index of the contact voltage within the window: ； in, These are the weighting coefficients; , These represent the maximum and minimum voltage values ​​within the sliding window, respectively. S52. When the sliding window simultaneously satisfies the stability index Contact current And the average window voltage When, mark as a candidate endpoint And verify that the above conditions are consistently met for the minimum stay time. ;in, The preset stability index; Indicates the voltage tolerance of the disconnect platform; S53, If the voltage exceeds the threshold again after the candidate termination point. If the arc initiation conditions are met, it is identified as a secondary arcing event caused by a bounce, and the segmented merging rule is applied to handle events where the interval time is less than the merging threshold. Arc ignition event.

7. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 6, characterized in that, Step S53 specifically includes: S531. After the candidate end point marked in step S52, continue monitoring backwards for a preset length. If the contact voltage does not exceed the threshold again within the sampling point sequence, then... If the arc initiation condition in step S4 is not met, then the candidate termination point is confirmed as the primary arc termination point. ; S532. If the voltage exceeds the threshold again within the monitoring range. If the arc initiation condition of step S4 is met, it is identified as a secondary arcing event caused by a bounce, and steps S51 and S52 are repeated for the secondary arcing event to determine its corresponding candidate end point. S533. Calculate the time interval gap between the end point of the first arc and the start point of the second arc. If gap is less than the merging threshold... If the two arcs are merged into a single arc event, the end point of the primary arc is updated to the end point of the secondary arc; if If it does not occur, it remains an independent arcing event, and the end point of the main arcing event is still the end point of the first arcing event. S534, Output the starting point of the main arcing event. and end point The starting point is determined in step S4. The endpoint is the final result after processing steps S531 to S533, and the start and end points of each secondary arcing event are output simultaneously.

8. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 7, characterized in that, Step S6 includes: S61. Generate an arc state mask based on the determined arc start and end points. ,in, This indicates that it is in an arcing state. This indicates that the circuit is in a non-arc state; S62, Define voltage bias Median of closed plateau voltage Define current bias This represents the median of the pre-trigger segment current. S63. Calculate the instantaneous power after bias correction. And the trapezoidal integral method is used only for Energy integration is performed at the sampling points: ； in, This represents the energy integral of the sampling point under arcing conditions. The sampling interval is denoted as .

9. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 8, characterized in that, Step S6 also includes: Based on arc state mask Calculate the total arcing time: ； Among them, for all The time intervals corresponding to the sampling points are accumulated; Based on the main arc initiation point determined in step S534 and the end point of the main combustion arc Calculate the main arc time .

10. The relay arcing parameter identification method based on high sampling rate and multi-resolution according to claim 1, characterized in that, Also includes: The identification results from steps S1 to S6 are scored for quality and a reason code is generated. After automatically adjusting the corresponding algorithm parameters based on the cause code category, re-collection is triggered.

Images