Fuse monitoring system of high-voltage connector and vehicle electrical performance monitoring system

Abstract

The invention relates to the technical field of fuse monitoring, in particular to a fuse monitoring system of a high-voltage connector and a vehicle electrical performance monitoring system, and the system comprises a data analysis module, a feature extraction module, a threshold generation module, a risk assessment module and a state monitoring module. According to the method, the continuous state quantity is constructed through serialization processing of the temperature change, the stress change and the resistance change, the degradation track obtains distinguishable expression through normalization correlation of the gradient characteristic and the change rate characteristic, and the display degree of the degradation trend in the multi-source data is enhanced through a principal component extraction mode. A response boundary capable of synchronously changing along with the environment and the state is formed in a weighted threshold and amplitude correction mode, and pre-judgment on early degradation and abnormal evolution is realized through hysteresis analysis and migration trend identification, so that the state description precision of the fuse element is improved, and the stability and foresight of vehicle electrical performance monitoring are enhanced.

Description

Technical Field

[0001] This invention relates to the field of fuse monitoring technology, and in particular to fuse monitoring systems for high-voltage connectors and vehicle electrical performance monitoring systems. Background Technology

[0002] The field of fuse monitoring technology belongs to the power system protection and condition detection direction. This field mainly focuses on the real-time status identification of fuses in circuits, the determination of circuit breaks under overcurrent conditions, the detection of the continuity integrity of high-voltage connection parts, and the perception of the operating status of the internal power system of vehicles or equipment. By monitoring the changes in current, voltage, temperature rise and conductive structure, the operating conditions of fuse elements can be judged. This field as a whole involves fuse structural characteristics, circuit anomaly discrimination logic, and condition sensing solutions under high-voltage or complex power supply environments.

[0003] Among them, the traditional high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system refer to the technical solutions used to identify the status of fuses and related electrical parameters at high-voltage connection locations or inside the vehicle's power system. Typically, it is used to determine whether the fuse is in an intact or open state by arranging sampling wires near the fuse, detecting resistance, or installing mechanically triggered fuse indicators. It also uses current sampling coils, voltage sampling nodes, or temperature-sensitive elements deployed on the vehicle's power supply lines to monitor the electrical performance of the lines. Traditional solutions mostly rely on comparing the potential difference before and after the fuse, detecting the continuity of whether the conductor has formed an open circuit, and recording the voltage fluctuations or current deviations in the vehicle's power supply circuit to complete the monitoring function.

[0004] Existing technologies rely on comparing the potential difference before and after the fuse, detecting conductor continuity, and single-point measurement signals such as temperature or current. The monitoring process is sensitive to transient disturbances and it is difficult to obtain synchronous identification of temperature changes, mechanical stress, and conduction quality at high-voltage connection points. The weak data correlation makes it difficult to identify abnormal evolution in advance. The single threshold judgment method is insufficient to respond to slow degradation characteristics and is prone to identification lag in the early aging stage of fuse elements. The electrical deviation of the vehicle power supply system under complex loads is difficult to be effectively distinguished, resulting in delayed risk warning, rough condition assessment, and difficulty in accurately capturing the life decay trend. Summary of the Invention

[0005] To address the technical problems existing in the prior art, embodiments of the present invention provide a fuse monitoring system for high-voltage connectors and a vehicle electrical performance monitoring system. The technical solutions are as follows:

[0006] On the one hand, it provides a fuse monitoring system for high-voltage connectors and a vehicle electrical performance monitoring system, which includes:

[0007] The data analysis module collects temperature, stress, and resistance data of the high-voltage connector and performs sequence smoothing. It calculates the temperature gradient, stress gradient, and resistance change rate on the smoothed data, generates state change data, and transmits it to the feature extraction module.

[0008] The feature extraction module extracts temperature gradient, stress gradient and resistance change rate based on the state change data, normalizes them and performs difference calculation to obtain state change input, extracts the principal component of degradation trend and performs classification judgment on the state change input, generates degradation feature coefficients and passes them to the threshold generation module.

[0009] The threshold generation module obtains the stress gradient and resistance change rate and combines them with the degradation characteristic coefficient to weighted analyze the initial response threshold, calls the temperature gradient data for amplitude correction, generates the dynamic response threshold, and transmits it to the risk assessment module.

[0010] The risk assessment module extracts temperature gradient, stress gradient and resistance change rate based on the state change data and calculates response hysteresis. It performs amplitude comparison on the dynamic response threshold and analyzes migration trend factors. It also makes risk judgments on temperature gradient, stress gradient and resistance change rate and generates vehicle electrical performance monitoring results.

[0011] The condition monitoring module performs refined and graded processing based on the vehicle electrical performance monitoring results to obtain the fuse condition monitoring results. It then performs life decay trend analysis and feature clustering operations on the fuse condition monitoring results to generate optimized vehicle electrical performance monitoring results.

[0012] As a further aspect of the present invention, the state change data includes a temperature change vector, a stress change vector, and a resistance change vector; the degradation characteristic coefficients include principal component coefficients, classification discrimination coefficients, and normalized difference coefficients; the dynamic response thresholds include stress correction thresholds, resistance correction thresholds, and temperature amplitude thresholds; the vehicle electrical performance monitoring results include a risk level index, a response hysteresis index, and a migration trend factor; and the optimized vehicle electrical performance monitoring results include a graded state index, a lifespan decay index, and a feature clustering index.

[0013] As a further aspect of the present invention, the data analysis module includes:

[0014] The data stream receiving submodule collects the temperature, stress and resistance data of the fuse of the high-voltage connector, integrates them in chronological order, judges the acquisition error by comparing point by point, performs error elimination operation, and performs segment connection on the remaining sequence after elimination to generate the original monitoring sequence.

[0015] The sequence smoothing submodule, based on the original monitoring sequence, calls the temperature, stress and resistance sequences, calculates the sequence volatility according to the difference values ​​of adjacent sampling points, and performs segment value replacement and segment mean reconstruction according to the volatility threshold to obtain a smoothed monitoring sequence.

[0016] The state feature submodule, based on the smoothed monitoring sequence, calls the temperature, stress, and resistance sequences, calculates the temperature gradient and stress gradient, analyzes the resistance change rate based on the ratio of the resistance sequence before and after, and combines the temperature gradient, stress gradient, and resistance change rate to generate state change feature data.

[0017] As a further aspect of the present invention, the fluctuation threshold is determined by calculating the absolute value of the difference values ​​of the temperature sequence, stress sequence and resistance sequence in chronological order through the distribution of the difference values ​​of adjacent sampling points in the statistical sequence, sorting all the absolute values ​​of the difference values ​​according to their numerical values ​​to form a difference distribution sequence, and extracting the median value of the difference distribution sequence.

[0018] As a further aspect of the present invention, the feature extraction module includes:

[0019] The gradient calculation submodule extracts temperature gradient, stress gradient value and resistance change rate based on the state change feature data, obtains the corresponding time series and performs difference comparison, filters out segments with stable change amplitude, organizes the gradient parameters at the same time position in a numerical accumulation manner, and generates gradient dataset.

[0020] The normalization generation submodule calls the temperature gradient, stress gradient, and resistance change rate in the gradient dataset and performs scaling to calculate the normalized difference. It then filters the input parameters in the normalized difference by numerical comparison to generate the normalized input quantity of state change.

[0021] The principal component classification submodule obtains the time series distribution of multiple parameters based on the normalized input of state change, calculates the covariance of the multi-parameter series to determine the trend correlation and selects the dominant trend component as the principal component of degradation trend, performs interval classification on the normalized input of state change, and generates degradation feature coefficients.

[0022] As a further aspect of the present invention, the threshold generation module includes:

[0023] The gradient parameter submodule acquires the stress gradient and resistance change rate and performs corresponding time calibration. It uses numerical comparison to judge the data differences at multiple time points and eliminates abnormal jump points. Based on the data volume of the remaining segment, it performs sequence integration and synchronization to generate the stress-resistance joint sequence quantity.

[0024] The degradation coefficient submodule calls the stress-resistance joint sequence quantity, performs weighting on the multi-time point data in the stress-resistance joint sequence quantity in combination with the degradation characteristic coefficient, removes the weighted results that exceed the amplitude benchmark value, and performs sequence accumulation based on the remaining weighted values ​​to obtain the weighted response basic result;

[0025] The threshold correction submodule performs an amplitude comparison between the weighted response baseline result and the temperature gradient data at the same time point, calculates the deviation ratio between the amplitude value of the temperature gradient data and the preset standard amplitude as the correction amount, and superimposes it with the weighted response baseline result to generate a dynamic response threshold.

[0026] As a further aspect of the present invention, the amplitude reference value is determined by performing numerical distribution analysis on the data segment after weighting the degradation characteristic coefficients, obtaining the amplitude values ​​of all sampling points within the segment, sorting them, and calculating the median value of the sorted amplitude values.

[0027] As a further aspect of the present invention, the risk assessment module includes:

[0028] The response hysteresis submodule extracts the time series of temperature gradient, stress gradient and resistance change rate based on the state change data, obtains the peak time points of multiple sequences and performs time difference calculation, uses the difference between peak time points to analyze the response difference, and generates the response hysteresis.

[0029] The threshold comparison submodule calls the response hysteresis and the dynamic response threshold to perform amplitude segment comparison on the amplitude sequences of temperature gradient, stress gradient and resistance change rate, and performs amplitude difference calculation based on the difference between the amplitude sequences inside and outside the threshold segment to generate migration trend factor.

[0030] The migration interval submodule calls the migration trend factor and the dynamic response threshold to perform interval division determination on temperature gradient, stress gradient and resistance change rate, and performs segment assignment determination and risk category classification between migration trend factor value and interval boundary to generate vehicle electrical performance monitoring results.

[0031] As a further aspect of the present invention, the status monitoring module includes:

[0032] The risk segmentation submodule, based on the vehicle electrical performance monitoring results, performs segment reading on the risk category, extracts the boundary difference between the segment position of the risk category and the risk classification benchmark value for classification judgment, calls the vehicle electrical operation index corresponding to the risk category as the classification mapping quantity, and generates fuse status monitoring results.

[0033] The lifespan trend submodule calls the fuse status monitoring results, obtains the resistance change and temperature change, compares them with the risk level value, performs trend difference using the time series slope of the change sequence and the segment position of the risk level, analyzes the lifespan decay vector, and generates the lifespan decay trend.

[0034] The feature clustering submodule calls the electrical behavior feature sequence of the life decay trend and the fuse status monitoring result and calculates the feature distance, extracts the amplitude range of the two and calculates the difference, and performs clustering in combination with the life decay trend value to generate optimized vehicle electrical performance monitoring results.

[0035] The risk classification benchmark value is set by analyzing the distribution characteristics of electrical performance data, based on the long-term distribution characteristics of temperature, current, and voltage indicators, and by summing the mean of the long-term distribution characteristics and the standard deviation of three times the mean.

[0036] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0037] By serializing temperature, stress, and resistance changes to construct continuous state variables, and by normalizing and correlating gradient and rate of change features to obtain a distinguishable representation of degradation trajectories, the visibility of degradation trends in multi-source data is enhanced through principal component extraction. A response boundary that can change synchronously with the environment and state is formed through weighted thresholding and amplitude correction. Early degradation and abnormal evolution can be pre-judged through hysteresis analysis and migration trend identification, thereby improving the accuracy of fuse state characterization and enhancing the stability and foresight of vehicle electrical performance monitoring. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0039] Figure 1 This is a schematic diagram of the system of the present invention;

[0040] Figure 2 This is a schematic diagram of the system framework of the present invention;

[0041] Figure 3 This is a flowchart of the data analysis module in this invention;

[0042] Figure 4 This is a flowchart of the feature extraction module in this invention;

[0043] Figure 5This is a flowchart of the threshold generation module in this invention;

[0044] Figure 6 This is a flowchart of the risk assessment module in this invention;

[0045] Figure 7 This is a flowchart of the status monitoring module in this invention. Detailed Implementation

[0046] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0047] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0048] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0049] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0050] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0051] This invention provides a fuse monitoring system for high-voltage connectors and a vehicle electrical performance monitoring system, such as... Figure 1-2 The diagram shows a fuse monitoring system for high-voltage connectors and a vehicle electrical performance monitoring system. The system includes:

[0052] The data analysis module collects temperature, stress, and resistance data of the high-voltage connector and performs sequence smoothing. It calculates the temperature gradient, stress gradient, and resistance change rate on the smoothed data, generates state change data, and transmits it to the feature extraction module.

[0053] The feature extraction module extracts temperature gradient, stress gradient and resistance change rate based on state change data, normalizes them and performs difference calculation to obtain state change input, extracts the principal component of degradation trend and performs classification judgment on state change input, generates degradation feature coefficients and passes them to the threshold generation module.

[0054] The threshold generation module obtains the stress gradient and resistance change rate and combines them with the degradation characteristic coefficient to weighted analyze the initial response threshold. It then calls the temperature gradient data for amplitude correction, generates the dynamic response threshold, and passes it to the risk assessment module.

[0055] The risk assessment module extracts temperature gradient, stress gradient and resistance change rate based on state change data and calculates response hysteresis. It performs amplitude comparison on dynamic response threshold and analyzes migration trend factors. It also makes risk judgments on temperature gradient, stress gradient and resistance change rate and generates vehicle electrical performance monitoring results.

[0056] The condition monitoring module performs refined and graded processing based on the vehicle electrical performance monitoring results to obtain the fuse condition monitoring results. It then performs life decay trend analysis and feature clustering operations on the fuse condition monitoring results to generate optimized vehicle electrical performance monitoring results.

[0057] The state change data includes temperature change vector, stress change vector, and resistance change vector; the degradation characteristic coefficients include principal component coefficients, classification discriminant coefficients, and normalized difference coefficients; the dynamic response thresholds include stress correction thresholds, resistance correction thresholds, and temperature amplitude thresholds; the vehicle electrical performance monitoring results include risk level index, response hysteresis index, and migration trend factor; and the optimized vehicle electrical performance monitoring results include graded state index, life decay index, and feature clustering index.

[0058] Specifically, such as Figure 2 , 3 As shown, the data analysis module includes:

[0059] The data stream receiving submodule collects the temperature, stress and resistance data of the fuse of the high-voltage connector, integrates them in chronological order, judges the acquisition error by comparing point by point, performs error elimination operation, and performs segment connection on the remaining sequence after elimination to generate the original monitoring sequence.

[0060] Temperature, stress, and resistance data of the fuses in the high-voltage connector are collected and monitored in real time by sensors. Temperature data is in degrees Celsius (°C), stress data in megapascals (MPa), and resistance data in ohms (Ω). The data is integrated chronologically at 1-second intervals to form a time series. Acquisition errors are assessed using a point-by-point comparison method: the difference between adjacent sampling points is calculated, and points with a temperature difference exceeding 2°C, a stress difference exceeding 0.5 MPa, or a resistance difference exceeding 0.1 Ω are identified as error points. Error removal is then performed: the data points corresponding to the error points are removed. After removal, the remaining sequence is segmented: the data segments before and after the removed points are connected, and the gaps are filled using linear interpolation to generate the original monitoring sequence. During the operation of the high-voltage connector, the following time-series data were collected: at time point t=1s, the temperature was 25°C, the stress was 10.0 MPa, and the resistance was 5.0 Ω; at t=2s, the temperature was 27.5°C, the stress was 10.6 MPa, and the resistance was 5.2 Ω; at t=3s, the temperature was 24.0°C, the stress was 9.8 MPa, and the resistance was 4.7 Ω. The temperature difference between t=2s and t=3s was calculated to be |27.5-24.0|=3.5°C, which exceeded the threshold of 2°C. Therefore, t=3s was determined to be an error point and discarded. The remaining sequences t=1s and t=2s were joined together, and interpolation was used to fill the t=3s position with a temperature of 26.25°C (based on the average of t=1s and t=2s), a stress of 10.3 MPa, and a resistance of 5.1 Ω, generating the original monitoring sequence.

[0061] Threshold settings are as follows: The temperature threshold of 2°C is based on experimental data verification. 100 sets of normal data points were collected in the experiment, and the distribution of differences between adjacent points was calculated. 95% of the data points had a difference of less than 2°C; therefore, 2°C was set as the threshold. The stress threshold of 0.5 MPa was verified experimentally. The standard deviation of stress differences in the experimental data was 0.2 MPa, and setting the threshold to 0.5 MPa covers 99% of normal fluctuations. The resistance threshold of 0.1 Ω is based on resistor specifications. The experimentally tested resistance variation range was ±0.05 Ω, and a threshold of 0.1 Ω was set to tolerate slight fluctuations. Temperature data was collected using an integrated temperature sensor with a sampling frequency of 1 Hz and a range of 0-100°C; stress data was collected using a strain gauge with a range of 0-20 MPa; and resistance data was collected through online contact resistance monitoring with a range of 0-10 Ω. In the example, data at t=1s was collected directly; data at t=2s was collected and stored; and data at t=3s was collected and the difference was calculated for judgment. Experimental verification process: In a laboratory environment, the operation of a high-voltage connector was simulated, and 500 data points were collected. Error judgment threshold verification: 5% of the points had a temperature difference exceeding 2°C, 1% had a stress difference exceeding 0.5 MPa, and 3% had a resistance difference exceeding 0.1 Ω. After rejection, the sequence integrity was 99%. Examples of experimental data are shown in Table 1.

[0062] Table 1: Experimental Data Table for Data Stream Reception

[0063]

[0064] As shown in Table 1, the experimental data were used to verify the threshold setting. The advantage of this method is that outliers are directly removed by calculating the difference between adjacent points and determining the threshold, thus ensuring the reliability of the sequence.

[0065] The sequence smoothing submodule, based on the original monitoring sequence, calls the temperature, stress and resistance sequences, calculates the sequence volatility based on the difference values ​​of adjacent sampling points, and performs segment value replacement and segment mean reconstruction according to the volatility threshold to obtain a smoothed monitoring sequence.

[0066] Based on the original monitoring sequence, the temperature, stress, and resistance sequences are invoked, and the sequence volatility is calculated according to the differences between adjacent sampling points: the sum of the absolute differences between each data point and the previous and next points is calculated as the volatility value. The volatility unit is dimensionless, and the calculation method is: Volatility = |Current point value - Previous point value| + |Current point value - Next point value|. Segment value replacement is performed according to the volatility threshold: when the volatility exceeds the threshold, the point value is replaced with the average of the two points before and after it. Segment mean reconstruction is then performed: for the replaced sequence, the average value within each segment is calculated for every 3 points, and the sequence is reconstructed to obtain a smoothed monitoring sequence. The temperature, stress, and resistance sequences are invoked, and the sequence volatility is calculated according to the differences between adjacent sampling points: Temperature data in the original sequence: Point 1 = 25°C, Point 2 = 26.25°C, Point 3 = 26.0°C, Point 4 = 27.0°C. Calculate the volatility at point 2: |26.25-25| + |26.25-26.0| = 1.25 + 0.25 = 1.50. The volatility threshold is set to 1.2. Since the volatility at point 2 is 1.50, exceeding the threshold, the value for point 2 is replaced with (25+26.0) / 2 = 25.5°C. Reconstruct the segments: Segment 1 (points 1-3) mean = (25+25.5+26.0) / 3 = 25.5°C; Segment 2 (point 4) mean = 27.0°C. The smoothed sequence is 25.5°C and 27.0°C.

[0067] Threshold setting reference: The volatility threshold of 1.2 is calculated based on experimental data. The experiment collected 200 sets of smoothed sequence data, calculated the volatility distribution, and found that 90% of the data points had volatility less than 1.2. Therefore, a threshold of 1.2 was set. Calculation process: The average volatility in the experimental data was 0.8, and the standard deviation was 0.2. Setting the threshold to 1.2 covers 95% of the normal data. The volatility value is calculated from the sequence point values. For example, point 2 (26.25°C) comes from the original sequence, with the preceding point at 25°C and the following point at 26.0°C; the difference is directly calculated. In the example, point 1 (25°C), point 2 (26.25°C), and point 3 (26.0°C) all come from the data stream receiving module output. Parameter assignment: Temperature sequence value range 20-30°C, stress sequence value range 9-11 MPa, and resistance sequence value range 4.5-5.5 Ω. In volatility calculation, the point value calls the stored sequence value. Experimental Verification Process: The experiment used 100 points from the original sequence. Volatility threshold verification: 10% of the points had a volatility exceeding 1.2. After replacement, the sequence smoothness improved, and the segment mean reconstruction error was less than 0.5%. Experimental Data: Example of point value volatility calculation: Point 1 volatility 1.0 (normal), Point 2 volatility 1.5 (exceeding the threshold). The advantage of this method is that by calculating volatility and replacing the threshold, the segment mean is directly reconstructed, reducing sequence noise.

[0068] The state characteristics submodule calls the temperature, stress and resistance sequences based on the smoothed monitoring sequence, calculates the temperature gradient and stress gradient, analyzes the resistance change rate based on the ratio of the resistance sequence before and after, and combines the temperature gradient, stress gradient and resistance change rate to generate state change characteristic data.

[0069] Based on the smoothed monitoring sequence, the temperature, stress, and resistance sequences are retrieved, and the temperature gradient is calculated: Temperature gradient = (Current temperature - Previous temperature) / Time interval, 1 second, unit °C / s. The stress gradient is calculated: Stress gradient = (Current stress - Previous stress) / Time interval, unit MPa / s. The rate of change of resistance is analyzed based on the ratio of the resistance before and after points: Rate of change of resistance = (Current resistance / Previous resistance) - 1, unitless. The temperature gradient, stress gradient, and rate of change of resistance are combined: For each point, the product of these three is calculated as the characteristic data of the state change, unit (°C / s)·(MPa / s)·(unitless). In the smoothed sequence: Temperature: Point 1 = 25.5°C, Point 2 = 27.0°C; Stress: Point 1 = 10.3 MPa, Point 2 = 10.5 MPa; Resistance: Point 1 = 5.1 Ω, Point 2 = 5.0 Ω. Calculate the temperature gradient at point 2: (27.0-25.5) / 1 = 1.5°C / s; stress gradient: (10.5-10.3) / 1 = 0.2 MPa / s; resistance change rate: (5.0 / 5.1)-1 ≈ -0.0196. Combine: State characteristic data = 1.5 × 0.2 × (-0.0196) ≈ -0.00588. The temperature gradient is calculated using the smoothed sequence temperature values, for example, point 2 temperature 27.0°C, point 1 temperature 25.5°C, the difference is calculated. The resistance change rate is calculated using the resistance values, point 2 resistance 5.0 Ω, point 1 resistance 5.1 Ω, the ratio is calculated. In the example, the point values ​​come from the smoothing module output. Parameter assignment: temperature gradient range -5 to 5°C / s, stress gradient range -1 to 1 MPa / s, resistance change rate range -0.1 to 0.1. In the combined calculation, the parameter values ​​are directly multiplied. Baseline value setting: The baseline value for state feature data is 0. Based on experimental data, an abnormal state is determined when the absolute value of the feature data exceeds 0.01. Calculation process: The experiment collected 50 sets of normal data. Feature data distribution: 95% of the data have an absolute value less than 0.01. The baseline value is set to 0, and the abnormal interval is |value|>0.01. Experimental verification process: The experiment uses a smoothed sequence of 50 points to calculate feature data. For example, the feature data for point 2 is -0.00588, with an absolute value less than 0.01, thus it is considered normal. Complete experimental data: Feature data value distribution, with a mean of 0.002 and a standard deviation of 0.003. The advantage of this method is that it directly combines multidimensional parameters through gradient calculation and rate of change analysis, improving the accuracy of state monitoring.

[0070] Specifically, such as Figure 2 , 4 As shown, the feature extraction module includes:

[0071] The gradient calculation submodule extracts temperature gradient, stress gradient and resistance change rate based on state change feature data, obtains the corresponding time series and performs difference comparison, filters out segments with stable change amplitude, organizes gradient parameters at the same time position in a numerical accumulation manner, and generates gradient dataset.

[0072] Based on the state change feature data, temperature gradient, stress gradient values ​​and resistance change rate are extracted; the state change feature data are obtained from the aforementioned output, temperature gradient sequence [0.1, 0.15, 0.18, 0.22, 0.19, 0.17, 0.14, 0.16, 0.12, 0.11] °C / s, stress gradient sequence [0.02, 0.025, 0.03, 0.035, 0.028, 0.022, 0.018, 0.02, 0.015, 0.01] MPa / s, resistance change rate sequence [0.008, 0.012, 0.015, 0.018, 0.014, 0.01, 0.009, 0.011, 0.007, 0.006]. Perform difference comparison: At time point t=2, the temperature gradient difference |0.15-0.1|=0.05°C / s, the stress gradient difference |0.025-0.02|=0.005MPa / s, and the resistance change rate difference |0.012-0.008|=0.004. Filter stable ranges of change: Set thresholds of 0.1°C / s for temperature gradient, 0.05MPa / s for stress gradient, and 0.01 for resistance change rate (based on experiments, 95% of the stable ranges in the experimental data have change rates less than these thresholds); For the window t=1 to 3, the average difference (temperature gradient: (0.05+0.03) / 2=0.04°C / s) is below the threshold of 0.1°C / s, indicating stability; for t=4 to 6, the average difference is 0.06°C / s (below the threshold), indicating stability; for t=7 to 9, the average difference is 0.04°C / s (stable). Numerical accumulation: The cumulative temperature gradient from t=1 to 3 is 0.1 + 0.15 + 0.18 = 0.43°C / s, the cumulative stress gradient is 0.02 + 0.025 + 0.03 = 0.075 MPa / s, and the cumulative rate of change of resistivity is 0.008 + 0.012 + 0.015 = 0.035. The gradient dataset is [0.43°C / s, 0.075 MPa / s, 0.035].

[0073] Threshold settings are as follows: Temperature gradient threshold of 0.1°C / s was experimentally verified. 100 temperature gradient data points were collected, and the difference distribution was calculated. 90% of the point differences were less than 0.1°C / s (e.g., experimental point difference mean 0.08, standard deviation 0.01). The threshold was set to cover 95% of the normal data. Stress gradient threshold of 0.05 MPa / s was experimentally verified. The experimental data difference mean was 0.03 MPa / s, standard deviation 0.005. 0.05 was set as the upper limit. Resistance change rate threshold of 0.01 was experimentally verified. The experimental difference mean was 0.007, standard deviation 0.002. Calculation process: The difference was calculated using a simulated experimental sequence. For example, the average value of the point difference sequence [0.05, 0.08, 0.06] was 0.063, which is less than the 0.1 threshold, indicating stability. Data acquisition process: State change feature data is retrieved from the aforementioned output; gradient sequences are retrieved from the aforementioned module storage (the time point t corresponds to the previous output value); difference calculation is performed by subtracting the current value from the previous value and taking the absolute value, e.g., t=2 retrieves a value of 0.15°C / s and t=1 retrieves a value of 0.1°C / s. Experimental verification process: 50 sets of sequences were tested in the laboratory, and the stable segment screening threshold was verified: points with a temperature gradient difference exceeding 0.1°C / s accounted for 5%, points with a stress gradient exceeding 0.05MPa / s accounted for 3%, and points with a resistance change rate exceeding 0.01 MPa / s accounted for 4%. The sequence coverage after screening was 98%. Experimental data examples are shown in Table 2.

[0074] Table 2: Gradient Sequence Experimental Data Table

[0075]

[0076] As shown in Table 2, the experimental data were used to verify the threshold settings. The advantage of this approach is that it directly integrates gradients from the same segment through difference comparison and numerical accumulation, thereby improving the completeness of the dataset.

[0077] The normalization generation submodule calls the temperature gradient, stress gradient, and resistance change rate in the gradient dataset and performs scaling to calculate the normalized difference. It then filters the input parameters in the normalized difference by numerical comparison to generate the normalized input quantity of state change.

[0078] The gradient dataset is retrieved, containing cumulative temperature gradient, cumulative stress gradient, and cumulative resistance change rate; the units are °C / s, MPa / s, and unitless, respectively. The gradient dataset is obtained from the aforementioned output, with three segments of cumulative data: segment A [cumulative temperature gradient = 0.43°C / s, cumulative stress gradient = 0.075MPa / s, cumulative resistance change rate = 0.035], segment B [0.39°C / s, 0.068MPa / s, 0.029], and segment C [0.37°C / s, 0.062MPa / s, 0.026]. Execution scaling: The cumulative temperature gradient for the entire sequence is a minimum of 0.37°C / s and a maximum of 0.43°C / s. The normalized temperature gradient for segment A is (0.43-0.37) / (0.43-0.37)=1.0. Similarly, the cumulative stress gradient is a minimum of 0.062MPa / s and a maximum of 0.075MPa / s. The normalized stress gradient for segment A is (0.075-0.062) / (0.075-0.062)≈1.0. The cumulative resistance change rate is a minimum of 0.026 and a maximum of 0.035. The normalized resistance change rate for segment A is ≈(0.035-0.026) / (0.035-0.026)≈1.0. Similarly, for segment B: the normalized temperature gradient is ≈0.17, the stress gradient is ≈0.75, and the resistance change rate is ≈0.33. Calculate the normalized difference: Difference in segment A = |1.0-1.0| + |1.0-1.0| + |1.0-1.0| = 0.0; Difference in segment B ≈ |0.17-0.75| + |0.75-0.33| + |0.33-0.17| ≈ 0.58 + 0.42 + 0.16 ≈ 1.16. Numerical comparison and filtering: Difference threshold 0.1 (based on experiments, 85% of normal points in the experimental data have a difference less than 0.1), difference in segment A 0.0 ≤ 0.1, retained; difference in segment B 1.16 > 0.1, discarded. Generate normalized input for state change: retain normalized values ​​of segment A [1.0, 1.0, 1.0].

[0079] Threshold setting reference: The difference threshold is set to 0.1. Through experimental calculation, 50 sets of normalized data were collected, and the difference distribution was calculated. The mean was 0.05, and the standard deviation was 0.02. The threshold of 0.1 was set to cover 90% of the normal data. Calculation process: The experimental sequence difference [0.02, 0.08, 0.12] had an average of 0.073, which was less than 0.1 and was therefore retained. Data acquisition process: The gradient dataset calls the aforementioned output sequence; the normalization calculation calls the global minimum and maximum values ​​(obtained from the sequence storage); the difference calculation calls the sum of the absolute values ​​of the subtraction of normalized values. In the example, the minimum and maximum values ​​are directly calculated from the sequence value, such as the cumulative temperature gradient sequence [0.37, 0.39, 0.43]. Parameter assignment: The normalized value range is 0-1. Experimental verification process: The experiment used a gradient set of 30 points. Difference threshold verification: Points exceeding 0.1 accounted for 15%, and the input coverage rate was 95% after filtering. Experimental data: Example of difference calculation, normalized value of segment C [≈0.0, 0.0, 0.0], difference values ​​0.0 ≤ 0.1 are retained. The advantage of this method is that it directly filters consistent input values ​​by scaling and comparing differences, reducing the impact of noise.

[0080] The principal component classification submodule obtains the time series distribution of multiple parameters based on the normalized input of state change, calculates the covariance of the multi-parameter series to determine the trend correlation and selects the dominant trend component as the principal component of the degradation trend, performs interval classification on the normalized input of state change, and generates degradation feature coefficients.

[0081] The normalized input for state change is called, which is obtained from the aforementioned output. The sequence points are: t1[1.0, 1.0, 1.0], t2[0.9, 0.9, 0.9], t3[0.8, 0.8, 0.8] (actual sequence). The time series distribution is obtained as follows: parameter index 1 (temperature) sequence [1.0, 0.9, 0.8], index 2 (stress) sequence [1.0, 0.9, 0.8], index 3 (resistance) sequence [1.0, 0.9, 0.8]. Calculate the trend correlation: Calculate the trend slope of index 1 sequence (least square method), point value [(1, 1.0), (2, 0.9), (3, 0.8)], slope = [Σ(xy)-ΣxΣy / n] / [Σx²-(Σx)² / n], where x is the index value and y is the point value; calculate Σx=6, Σy=2.7, Σxy=1×1.0+2×0.9+3×0.8=1.0+1.8+2.4=5.2, n=3, Σx²=14, slope = (5.2-6×2.7 / 3) / (14-36 / 3)=(5.2-5.4) / (14-12)=(-0.2) / 2=-0.1; similarly, the slopes of indices 2 and 3 are both -0.1. Screening for dominant trend components: All slopes with an absolute value of 0.1 are less than the threshold of 0.5. The component with the largest absolute value, index 1 (slope -0.1), is selected as the dominant trend component, with a value of -0.1. As the principal component for the degradation trend: value -0.1. Performing interval classification: Interval 0-0.3 corresponds to label 1 (low degradation), 0.3-0.6 to label 2 (medium degradation), and 0.6-1.0 to label 3 (high degradation); principal component values ​​with an absolute value of 0.1 in the 0-0.3 interval are labeled as label 1. Generating degradation feature coefficients: coefficient = 1 × (-0.1) = -0.1.

[0082] Threshold setting reference: The trend correlation threshold is set to 0.5. Through experimental calculation, 50 sets of slope data were tested. The absolute value distribution had a mean of 0.3 and a standard deviation of 0.1. Setting it to 0.5 covers 95% of significant trends. Calculation process: The experimental slope sequence [0.2, -0.3, 0.4] had a maximum absolute value of 0.4, which was less than 0.5 and was retained as a component. Interval classification setting: Based on the distribution of degenerate values ​​in the experimental data, values ​​<0.3 were considered low risk (label 1). Data acquisition process: Normalized input values ​​were called from the aforementioned sequence; trend correlation calculation used sequence point values ​​and time indices; classification used principal component values ​​and preset intervals for comparison. In the example, slope calculation directly used point values ​​and index values. Parameter assignment: The trend slope range was -1 to 1. Experimental verification process: The experiment used 20 input points. Interval classification verification: Label 1 interval covered 65% of normal data, Label 2 covered 30%, and Label 3 covered 5%. Experimental data: Example of sequence point calculation, index 1 sequence [1.0, 1.0, 1.0] with a slope of 0 (no trend). The advantage of this method is that it directly generates quantified feature coefficients through trend correlation calculation and interval classification, thereby improving the accuracy of state representation.

[0083] Specifically, such as Figure 2 , 5 As shown, the threshold generation module includes:

[0084] The gradient parameter submodule acquires the stress gradient and resistance change rate and performs corresponding time calibration. It uses numerical comparison to judge the data differences at multiple time points and eliminates abnormal jump points. Based on the data volume of the remaining segment, it performs sequence integration and synchronization to generate the stress-resistance joint sequence quantity.

[0085] Based on state change characteristic data, this includes a stress gradient sequence (unit: MPa / s) and a resistance change rate sequence (unitless). The time series is retrieved from the aforementioned output, with a sampling interval of 1 second and a sequence length of 5 points. Time alignment is performed: the alignment method checks the consistency of timestamps at each time point (timestamps are obtained from the sensor acquisition system, unit: seconds) to ensure time point alignment; the sampling rate is set to 1 second, and time point indices 1 to 5 correspond to times 0s, 1s, 2s, 3s, and 4s, respectively, and the sequences must be completely aligned. Numerical comparison is used to determine differences in data across multiple time points: the absolute difference between the stress gradient and resistance change rate at adjacent time points is calculated (difference calculation = |current point value - previous point value|), with the same units as the original sequence. Exclude abnormal jump points: Set difference thresholds, with a stress gradient difference threshold of 0.02 MPa / s (based on experimental data, 95% of normal differences are less than 0.02 MPa / s in experimental verification) and a resistance change rate difference threshold of 0.005 (90% of normal differences are less than 0.005 in experimental verification). If the difference exceeds the threshold, it is judged as an abnormal point and removed. Perform sequence integration and synchronization based on the remaining data volume: reconstruct the remaining point sequence, perform linear interpolation to fill the missing time points (interpolation = (previous point value + next point value) / 2), and generate a stress-resistance joint sequence (including stress gradient and resistance change rate aligned sequences). Obtain the stress gradient sequence [0.02, 0.025, 0.03, 0.035, 0.01] MPa / s and the resistance change rate sequence [0.005, 0.007, 0.006, 0.008, 0.002] from the state change feature data; time point indices 1-5 correspond to 0s, 1s, etc. Time calibration verifies index alignment. Numerical comparison: Stress gradient difference at time point 2 (index 2) |0.025-0.02|=0.005MPa / s (less than the threshold of 0.02, normal); Stress gradient difference at time point 5 (index 5) |0.01-0.035|=0.025MPa / s (exceeds the threshold of 0.02, anomaly removed); similar resistance change rate differences are all less than 0.005. After excluding anomalies, the remaining stress sequence points are [0.02, 0.025, 0.03, 0.035] (index 1-4), the resistance sequence is [0.005, 0.007, 0.006, 0.008], and index 5 is missing. Sequence integration and synchronization: Interpolated stress at index 5 = (0.035 + set value) / 2 (skip if no value is set, no interpolation is needed in this example); Reconstruct the sequence time indices 1-5, fully aligned, generating the stress-resistance joint sequence quantity: stress gradient [0.02, 0.025, 0.03, 0.035, null] MPa / s (null indicates a missing value), but in this example, the missing value at index 5 is not processed. After sequence integration and synchronization, the stress gradient is [0.02, 0.025, 0.03, 0.035, interpolation setting value], and the actual sequence has 5 complete values.Threshold settings are as follows: The stress gradient difference threshold is set at 0.02 MPa / s. Experimental calculations were performed, collecting 100 stress gradient data points and calculating the difference distribution (e.g., a point difference sequence [0.01, 0.015, 0.005] with an average difference of 0.01 MPa / s). 95% of the point differences were less than 0.02 MPa / s (experimental data showed a mean difference of 0.012 and a standard deviation of 0.003). Therefore, a threshold of 0.02 was set to cover normal fluctuations. The resistance change rate difference threshold is set at 0.005. Experimental verification showed a mean difference of 0.003 and a standard deviation of 0.001, so a threshold of 0.005 was set. Data acquisition process: The stress gradient sequence calls the stored values ​​from the aforementioned modules (e.g., directly calling index values); the difference calculation calls the current and previous values ​​(e.g., index 2 value 0.025 and index 1 value 0.02). Experimental verification process: The laboratory simulated the operation of a high-voltage connector, collecting 200 data points. Anomaly exclusion thresholds were verified: 5% of the data points had a stress gradient difference exceeding 0.02, and 10% had a resistance change rate exceeding 0.005. After these exclusions, the sequence integrity was 98%. Examples of experimental data are shown in Table 3.

[0086] Table 3: Experimental Data for Gradient Parameters

[0087]

[0088] As shown in Table 3, the experimental data were used to verify the threshold setting and sequence integration. The advantage of this approach is that it directly ensures sequence continuity through difference alignment and interpolation integration.

[0089] The degradation coefficient submodule calls the stress-resistance joint sequence quantity, performs weighting on the multi-time point data in the stress-resistance joint sequence quantity combined with the degradation characteristic coefficient, removes the weighted results that exceed the amplitude benchmark value, and performs sequence accumulation based on the remaining weighted values ​​to obtain the weighted response base result;

[0090] The system calls the combined stress-resistance sequence, which includes the stress gradient sequence and the resistance rate of change sequence (units MPa / s and unitless); the sequence length is 5 points (time indices 1-5). Degradation characteristic coefficients are obtained from the aforementioned principal component classification submodule (e.g., coefficient value -0.1). Weighting is performed: for each time point, the weighted result is calculated as: stress gradient value × degradation characteristic coefficient + resistance rate of change value × degradation characteristic coefficient, units are the same as the sequence. Weighted results exceeding the amplitude benchmark value are removed: the amplitude benchmark value is set to 0.5 (experimental setting), the absolute value of the weighted result is calculated, and if the absolute value > 0.5, the result at that point is removed. Sequence accumulation: the remaining weighted results are summed (cumulative sum = sum of remaining weighted results). The weighted response baseline result (numerical result) is obtained. The stress-resistance combined sequence is obtained from the aforementioned output: stress gradient sequence [0.02, 0.025, 0.03, 0.035, 0.01] MPa / s (index 1-5), resistance change rate sequence [0.005, 0.007, 0.006, 0.008, 0.002]; degradation characteristic coefficient call value -0.1. Perform weighted calculation: Index 1 weighted result = (0.02 × (-0.1)) + (0.005 × (-0.1)) = -0.002 + (-0.0005) = -0.0025; similarly, Index 2: (0.025 × (-0.1)) + (0.007 × (-0.1)) = -0.0025 + (-0.0007) = -0.0032; Index 3: -0.0030 + (-0.0006) = -0.0036; Index 4: -0.0035 + (-0.0008) = -0.0043; Index 5: -0.0010 + (-0.0002) = -0.0012. Amplitude baseline value 0.5: The absolute value of all weighted results is less than 0.5 (maximum 0.0043), with no rejection points. Sequence accumulation: The remaining weighted results [-0.0025, -0.0032, -0.0036, -0.0043, -0.0012] have a cumulative sum of -0.0148. The weighted response base result is -0.0148.

[0091] Threshold setting reference: The amplitude baseline value of 0.5 was calculated experimentally. The experiment tested 80 weighted data points with an absolute value distribution mean of 0.2 and a standard deviation of 0.1. Setting it to 0.5 covers 99% of normal data (e.g., in a weighted sequence [0.1, -0.3, 0.4], the maximum absolute value is 0.4, which is less than 0.5). Data acquisition process: The stress gradient sequence calls the aforementioned output values; the weighted calculation calls the stress gradient value, resistance change rate value, and degradation characteristic coefficient value (coefficients call stored values); the amplitude calculation calls the absolute value of the weighted result. In the example, the weighted value is calculated directly. Experimental verification process: The experiment used a 50-point sequence. Baseline value verification: points exceeding 0.5 accounted for 1%, and the cumulative result stability was 98%. The advantage of this method is that consistent response values ​​are directly accumulated through weighting and amplitude filtering.

[0092] The threshold correction submodule performs an amplitude comparison between the weighted response baseline result and the temperature gradient data at the same time point, calculates the deviation ratio between the amplitude value of the temperature gradient data and the preset standard amplitude as the correction amount, and superimposes it with the weighted response baseline result to generate a dynamic response threshold.

[0093] Based on the weighted response baseline result, a value such as -0.0148 and temperature gradient data are used; values ​​at the same time point are retrieved (e.g., values ​​corresponding to the index). Amplitude comparison is performed: the difference is calculated as |weighted response baseline result - temperature gradient value|, with units based on parameters (the numerical difference has no unit, it is actually a proportional value). Correction amount determination: based on the temperature gradient data amplitude range (e.g., -1 to 1°C / s), the average amplitude is calculated (average = Σ|temperature gradient value| / sequence length), and the correction amount is calculated as (difference × average amplitude) / reference value (the reference value is set to 1, fixed). Overlay: the dynamic response threshold = weighted response baseline result + correction amount, generating the final threshold result. The weighted response baseline result is retrieved as -0.0148; the temperature gradient data sequence is [0.01, 0.012, 0.015, 0.018, 0.01]°C / s (indexes 1-5); the time points correspond to the same index. Amplitude Comparison: Index 1 Difference = |-0.0148-0.01| = 0.0248; Index 2: |-0.0148-0.012| = 0.0268; Index 3: 0.0298; Index 4: 0.0328; Index 5: 0.0248. Correction Calculation: Temperature gradient amplitude range is calculated by summing the absolute values ​​of the sequences, average amplitude = (|0.01| + |0.012| + |0.015| + |0.018| + |0.01|) / 5 = 0.0132°C / s; Reference value setting 1 (fixed baseline). Correction amount calculation: Index 1 correction amount = (0.0248 × 0.0132) / 1 ≈ 0.000327; Index 2: (0.0268 × 0.0132) / 1 ≈ 0.000354; Index 3: (0.0298 × 0.0132) / 1 ≈ 0.000393; Index 4: (0.0328 × 0.0132) / 1 ≈ 0.000433; Index 5: (0.0248 × 0.0132) / 1 ≈ 0.000327. Overlay: Index 1 dynamic response threshold = -0.0148 + 0.000327 ≈ -0.014473; similar to Index 2: -0.0148 + 0.000354 ≈ -0.014446; Index 3: -0.0148 + 0.000393 ≈ -0.014407; Index 4: -0.0148 + 0.000433 ≈ -0.014367; Index 5: -0.0148 + 0.000327 ≈ -0.014473. Generate a dynamic response threshold sequence [-0.014473, -0.014446, -0.014407, -0.014367, -0.014473]. Threshold setting reference: The temperature gradient amplitude range is obtained through experiments, with experimental data ranging from [-0.5, 0.5] °C / s; the average amplitude calculation calls the mean of the absolute values ​​of the sequence (e.g., the average amplitude of the experimental sequence [0.1, -0.2] is 0.15); reference value 1 is fixed.Data acquisition process: The weighted response base result calls the stored value; the temperature gradient value calls the output of the gradient calculation module; the difference calculation calls both values; the correction calculation calls the difference, average amplitude, and reference value. In the example, the parameter values ​​are directly called. Experimental verification process: The experiment uses 50 data points, and the range verification covers 95% of the scenarios. The advantage of this method is that it directly generates accurate thresholds through amplitude comparison and correction superposition.

[0094] Specifically, such as Figure 2 , 6 As shown, the risk assessment module includes:

[0095] The response hysteresis submodule extracts the time series of temperature gradient, stress gradient and resistance change rate based on state change data, obtains the peak time points of multiple series and performs time difference calculation, uses the difference between peak time points to analyze response difference, and generates response hysteresis.

[0096] Based on state change characteristic data, it includes temperature gradient sequences (unit: °C / s), stress gradient sequences (unit: MPa / s), and resistance change rate sequences (unitless); the sequence length is 5 points (time indices 1-5 correspond to times 0s, 1s, 2s, 3s, and 4s). Sequence extraction: call the stored values, for example, temperature gradient sequence [0.01, 0.012, 0.015, 0.018, 0.01] °C / s, stress gradient sequence [0.02, 0.025, 0.03, 0.035, 0.01] MPa / s, and resistance change rate sequence [0.005, 0.007, 0.006, 0.008, 0.002]. Obtain the peak time points of multiple sequences: The peak is defined as the maximum value point in the sequence; calculate the maximum value position for each sequence. The temperature gradient peak is at index 4 (value 0.018°C / s), the stress gradient peak is at index 4 (value 0.035MPa / s), and the resistance change rate peak is at index 4 (value 0.008). Perform time difference calculation: calculate the difference between peak time points, difference = |temperature peak index - stress peak index| × time interval (1 second), unit is seconds; for example, the difference between the temperature and stress peak indices |4-4|=0, the difference is 0 seconds; the difference between the temperature and resistance peak indices |4-4|=0 seconds; the difference between the stress and resistance peak indices |4-4|=0 seconds. Analyze the response difference using the difference between peak time points: response difference = sum of all differences / number of differences; calculate the number of differences 3 (temperature-stress, temperature-resistance, stress-resistance), response difference = (0+0+0) / 3=0 seconds. Generate the response hysteresis (numerical result 0 seconds). Example: The sequence is retrieved from the aforementioned output; the peak time point is obtained by comparing the maximum value of the retrieved sequence value, such as the temperature sequence value [0.01, 0.012, 0.015, 0.018, 0.01], where the maximum value of 0.018 is at index 4. Time difference calculation involves subtracting the index values, taking the absolute value, and multiplying by the time interval (fixed at 1 second). Response difference calculation involves summing the differences and dividing by the quantity. In the example, the index values ​​are directly retrieved. Threshold setting reference: No threshold is directly involved, but the difference analysis is based on the index difference; in experimental verification, the index difference range is 0-2 seconds (a reasonable range). Data acquisition process: Sequence values ​​are retrieved and stored; peak positions are obtained by comparing sequence point values ​​(e.g., index 4 value 0.018 is greater than others); difference calculation involves retrieving the index values. Experimental verification process: 100 sets of sequences were tested in the laboratory. Peak time difference verification: 90% of the index difference points were 0 seconds, the average response difference was 0.05 seconds, and the standard deviation was 0.01 seconds; experimental data examples are shown in Table 4.

[0097] Table 4: Experimental Data on Response Lag

[0098]

[0099] As shown in Table 4, the experimental data were used to verify the acquisition of peak time points and the calculation of time differences. The advantage of this approach is that it directly quantifies the lag effect through peak time difference calculation and response difference analysis.

[0100] The threshold comparison submodule calls the response hysteresis and dynamic response threshold to perform amplitude segment comparison on the amplitude sequences of temperature gradient, stress gradient and resistance change rate. It uses the difference between the amplitude sequences inside and outside the threshold segment to perform amplitude difference calculation and generate migration trend factor.

[0101] Call the response hysteresis and dynamic response threshold; sequence length 5 points (time index 1-5). Perform amplitude segment comparison on the amplitude sequences of temperature gradient, stress gradient and resistance change rate: the amplitude sequence is defined as the absolute value sequence of each sequence (unit is the same as the original sequence); temperature gradient amplitude sequence [0.01, 0.012, 0.015, 0.018, 0.01] °C / s, stress gradient amplitude sequence [0.02, 0.025, 0.03, 0.035, 0.01] MPa / s, resistance change rate amplitude sequence [0.005, 0.007, 0.006, 0.008, 0.002]. The amplitude range is defined as the range of the absolute value sequence of the dynamic response threshold (e.g., absolute values ​​of the threshold sequence [0.014473, 0.014446, 0.014407, 0.014367, 0.014473]). Within the range: the amplitude value is within ±0.001 of the absolute value of the threshold (experimental setting); outside the range: it exceeds this range. Amplitude difference calculation is performed using the difference between the amplitude sequence inside and outside the threshold range: Difference = |Amplitude Value - Absolute Value of Threshold|; if within the range, the difference is included in the inner range difference; if outside the range, the difference is included in the outer range difference; calculate the average inner range difference and the average outer range difference; Amplitude Difference = Average Outer Range Difference - Average Inner Range Difference (units are not specified). Generate a migration trend factor (numerical result). Example: Response lag calls a value of 0 seconds; dynamic response threshold sequence calls a stored value; amplitude sequence calculation calls the absolute value of the original sequence (e.g., the absolute value of the temperature gradient sequence value is taken after calling). Amplitude range comparison: Index 1 temperature amplitude 0.01, threshold absolute value 0.014473, difference |0.01-0.014473|=0.004473; range threshold absolute value ±0.001=[0.013473, 0.015473], 0.01 is not within the range (outer zone); similarly, Index 4 temperature amplitude 0.018 is within [0.013367, 0.015367] (inner zone). Amplitude difference calculation: Inner zone difference point index 4 value |0.018-0.014367|=0.003633, average value 0.003633 (only one point); Outer zone difference point index 1 value 0.004473, index 2 value |0.012-0.014446|=0.002446, index 3 value |0.015-0.014407|=0.000593, index 5 value |0.01-0.014473|=0.004473, average value (0.004473+0.002446+0.000593+0.004473) / 4≈0.002996; Amplitude difference = 0.002996-0.003633≈-0.000637. The migration trend factor is -0.000637.

[0102] Threshold setting reference: The absolute value of the threshold within the segment range is ±0.001. Through experimental calculation, 50 sets of data were tested. 85% of the amplitude values ​​were within ±0.001 of the threshold (the experimental amplitude sequence [0.014, 0.015] had a threshold of 0.0145, and the difference was 0.0005, which was within the range). Calculation process: The mean difference of the experimental sequence was 0.0003, and ±0.001 was set to cover 90% of normal fluctuations. Data acquisition process: The amplitude sequence calls the absolute value of the original sequence; the segment comparison calls the absolute value of the threshold (calculated from the dynamic response threshold sequence); the difference calculation calls the amplitude value and the absolute value of the threshold. In the example, the parameter values ​​are directly called. Experimental verification process: The experiment used 30 points in the sequence. Segment range verification: 70% of the points were within the range, the mean amplitude difference was -0.0005, and the standard deviation was 0.0002. The advantage of this method is that it directly generates a trend quantification factor through the calculation of differences within and outside the segment.

[0103] The migration interval submodule calls the migration trend factor and dynamic response threshold to perform interval division judgment on temperature gradient, stress gradient and resistance change rate. It uses the migration trend factor value and interval boundary to perform segment assignment judgment and risk category classification, and generates vehicle electrical performance monitoring results.

[0104] Invoke the migration trend factor and dynamic response threshold; sequence length 5 points. Perform interval division judgment on temperature gradient, stress gradient, and resistance change rate: the interval is based on the dynamic response threshold sequence value range (e.g., minimum value -0.014473, maximum value -0.014367), dividing into three intervals: low zone (threshold < -0.01442), medium zone (-0.01442 ≤ value ≤ -0.01440), and high zone (value > -0.01440); the boundary value is calculated as the sequence mean ± standard deviation (experimental settings). Perform segment assignment judgment between the migration trend factor value and the interval boundary: assignment judgment = migration trend factor value compared with the interval boundary; if the value < low zone boundary, it belongs to the low zone; otherwise, if the value ≤ high zone boundary, it belongs to the medium zone; otherwise, it belongs to the high zone. Risk category classification: low zone corresponds to risk category 1 (low risk), medium zone to category 2 (medium risk), and high zone to category 3 (high risk). Generate vehicle electrical performance monitoring results (risk category values). Example: Migration trend factor call value -0.000637; Dynamic response threshold sequence call [-0.014473, -0.014446, -0.014407, -0.014367, -0.014473]. Interval division determination: Average calculation (-0.014473-0.014446-0.014407-0.014367-0.014473) / 5≈-0.0144332; Standard deviation calculation: Sum of squared differences / square root of 4≈0.000044 (set), low boundary <-0.01442 (average -0.0000132), middle zone -0.01442 to -0.01440 (average +0.0000332), high zone >-0.01440. Classification: The migration trend factor value of -0.000637 is compared to the boundary; -0.000637 > -0.01442, classifying it as low-risk. Risk category classification: Low-risk zone corresponds to category 1. Generate vehicle electrical performance monitoring result 1.

[0105] Threshold setting reference: Interval boundaries are based on sequence statistics, with the low-zone boundary set at half the standard deviation of the mean (experimental sequence standard deviation 0.000044, half the standard deviation 0.000022, boundary -0.0144332-0.000022≈-0.0144552, simplified -0.01442); Calculation process: Experimental data sequence mean -0.0144, standard deviation 0.00005, boundary setting covers 80% of the data. Data acquisition process: Interval division calls the threshold sequence value to calculate the mean and standard deviation; attribution judgment calls the migration trend factor value and compares it with the boundary. In the example, the mean and standard deviation are calculated directly. Experimental verification process: The experiment uses 50 data points, interval division verification: low zone points account for 60%, middle zone 30%, high zone 10%. The advantage of this method is that it directly outputs monitoring results through interval attribution and risk classification.

[0106] Specifically, such as Figure 2 , 7 As shown, the status monitoring module includes:

[0107] The risk segmentation submodule, based on the vehicle electrical performance monitoring results, performs segment reading for risk categories, extracts the boundary difference between the segment position of the risk category and the risk classification benchmark value for classification judgment, calls the vehicle electrical operation index corresponding to the risk category as the classification mapping quantity, and generates fuse status monitoring results.

[0108] Based on the vehicle electrical performance monitoring results (risk category value 1); retrieve the stored value. Perform segment reading for the risk category: the segment is defined as the storage location index of the risk category value (e.g., the category value is stored in register address 0x100); read the value 1 at this address. Extract the boundary difference between the segment position of the risk category and the risk classification benchmark value: the segment position is the risk category value itself (value 1); the risk classification benchmark value is set to 0.2 (experimental setting); boundary difference = |risk category value - benchmark value| = |1 - 0.2| = 0.8. Perform classification judgment: the classification rule is that if the boundary difference ≤ 0.5, it is classified as level 1; 0.5 < difference ≤ 1.0, it is classified as level 2; difference > 1.0, it is classified as level 3; 0.8 is in the range of 0.5-1.0, so it is classified as level 2. Retrieve the vehicle electrical operation index corresponding to the risk category: the index mapping relationship is that category 1 corresponds to index value 0.1, category 2 corresponds to 0.3, and category 3 corresponds to 0.5; level 2 retrieves index value 0.3. Classification mapping amount = classification value × index value = 2 × 0.3 = 0.6. Generate fuse status monitoring results (value 0.6). Example: Monitoring results call register value 1; baseline value 0.2 calls stored value; boundary difference calculation calls two values; grading judgment calls boundary difference value and interval comparison (interval boundaries 0.5 and 1.0); index value calls mapping table stored value; mapping quantity calculation calls grading value and index value. In the example, parameter values ​​are called directly.

[0109] Threshold setting reference: The risk grading baseline value is 0.2. Through experimental calculation, 50 sets of risk data were collected. The mean of the category value is 1.2, and the standard deviation is 0.3. Setting it to 0.2 covers 90% of scenarios (e.g., the mean of the category sequence [1, 1, 2] is 1.33, and the baseline of 0.2 ensures that the boundary difference is greater than 0.5). Interval boundary settings of 0.5 and 1.0: The mean of the boundary difference distribution is 0.8, and 0.5 and 1.0 are used to divide the data into three levels. Data acquisition process: Risk category value is called from the register; baseline value is called from storage; boundary difference calculation calls both values. Experimental verification process: The experiment uses 30 sets of data, with grading verification: 70% of the data are graded into two levels. The advantage of this method is that it directly quantifies the circuit breaker status through boundary difference grading and indicator mapping.

[0110] The lifespan trend submodule calls the fuse status monitoring results, obtains the resistance and temperature changes, compares them with the risk level values, performs trend difference analysis using the time series slope of the change sequence and the segment position of the risk level, analyzes the lifespan decay vector, and generates the lifespan decay trend.

[0111] Retrieve the fuse status monitoring result (value 0.6); obtain the current change sequence (unit: A / s), voltage change sequence (unit: V / s), and temperature rise change sequence (unit: °C / s); the sequence length is 5 points (time index 1-5). For example, the current change sequence is [0.5, 0.6, 0.7, 0.8, 0.9] A / s, the voltage change is [0.05, 0.06, 0.07, 0.08, 0.09] V / s, and the temperature rise change is [0.3, 0.32, 0.34, 0.36, 0.38] °C / s. Compare the trend with the risk level value: the risk level value is the fuse status value of 0.6. The time series slope of the changing sequence is calculated as follows: Slope = (End value - First value) / (Time interval × 4) (1 second interval); Current slope = (0.9 - 0.5) / 4 = 0.1 A / s², Voltage slope = (0.09 - 0.05) / 4 = 0.01 V / s², Temperature rise slope = (0.38 - 0.3) / 4 = 0.02 °C / s². Trend difference is performed with the risk level segment location: Segment location is defined as the interval where the risk value is located (risk value 0.6 is within the preset interval [0.4, 0.8]); Trend difference = |Slope mean - Risk value|, Slope mean = (0.1 + 0.01 + 0.02) / 3 ≈ 0.0433, Difference = |0.0433 - 0.6| ≈ 0.5567. Analyze the lifetime decay vector: Vector = Trend Difference × Risk Value = 0.5567 × 0.6 ≈ 0.334. Generate a lifetime decay trend (value 0.334).

[0112] Table 5: Experimental Data on Lifespan Trend

[0113]

[0114] As shown in Table 5, the experimental data were used for slope calculation and trend analysis. Threshold settings were based on the risk interval [0.4, 0.8], with a mean risk value distribution of 0.6, covering 80% of the data. Data acquisition process: The changing sequence retrieved sensor values; slope calculation retrieved the first and last point values; trend difference retrieved the mean slope and risk value. Experimental verification process: 20 sets of sequences were used in the experiment, and the slope calculation error was <5%. The advantage of this method is that it directly generates the attenuation amount by comparing the slope trend and the risk interval.

[0115] The feature clustering submodule calls the electrical behavior feature sequences of the life decay trend and the fuse status monitoring results and calculates the feature distance, extracts the amplitude range of the two and calculates the difference, and performs clustering in combination with the life decay trend value to generate optimized vehicle electrical performance monitoring results.

[0116] The lifespan degradation trend (value 0.334) and fuse condition monitoring results (value 0.6) are retrieved. The electrical behavior characteristic sequence is non-numerical (e.g., [fuse deformation, contact oxidation]), requiring quantification: Let fuse deformation = 0.8, contact oxidation = 0.6 (experimental quantification standard). Calculate the characteristic distance = |lifespan degradation value - characteristic quantification value|; for fuse deformation distance = |0.334 - 0.8| = 0.466, for contact oxidation distance = |0.334 - 0.6| = 0.266. Extract the amplitude ranges of both: the lifespan degradation amplitude range is defined as [0.3, 0.4] (experimental setting), and the characteristic quantification value amplitude range is [0.5, 0.7] (experimental setting). Calculate the difference = |median of lifespan interval - median of characteristic interval| = |0.35 - 0.6| = 0.25. Clustering is performed based on the lifespan decay trend value: the clustering rule is that if the distance is ≤0.3 and the difference is ≤0.2, then it belongs to cluster 1; otherwise, it belongs to cluster 2. For contact oxidation, the distance 0.266 ≤ 0.3 but the difference 0.25 > 0.2 belongs to cluster 2. Optimized vehicle electrical performance monitoring results are generated (cluster number 2). Example: lifespan decay value calls 0.334; fuse status value calls 0.6; feature sequence quantization calls the mapping table (fuse deformation → 0.8, contact oxidation → 0.6); distance calculation calls lifespan value and feature value; interval median calculation calls the interval boundary mean (lifespan interval [0.3, 0.4] median 0.35, feature interval [0.5, 0.7] median 0.6); cluster judgment calls distance value and difference.

[0117] Threshold settings referenced the following: Quantization standards were set experimentally, with fuse deformation corresponding to an 80% failure probability value of 0.8; Interval boundary experimental settings showed that lifespan decay values ​​were concentrated in [0.3, 0.4], and feature values ​​were concentrated in [0.5, 0.7]; Clustering thresholds of 0.3 and 0.2 were experimentally verified, with a mean distance of 0.25 and a mean difference of 0.22. Data acquisition process: Feature quantization invokes preset mapping; distance calculation invokes lifespan values ​​and feature values; interval median invokes boundary value calculation. Experimental verification process: 50 sets of data were tested, with a clustering accuracy of 92%. The advantage of this method is that it directly optimizes monitoring results through feature distance and interval difference.

[0118] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. 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 be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A fuse monitoring system for high-voltage connectors and a vehicle electrical performance monitoring system, characterized in that, The system includes: The data analysis module collects temperature, stress, and resistance data of the high-voltage connector and performs sequence smoothing. It calculates the temperature gradient, stress gradient, and resistance change rate on the smoothed data, generates state change data, and transmits it to the feature extraction module. The feature extraction module extracts temperature gradient, stress gradient and resistance change rate based on the state change data, normalizes them and performs difference calculation to obtain state change input, extracts the principal component of degradation trend and performs classification judgment on the state change input, generates degradation feature coefficients and passes them to the threshold generation module. The threshold generation module obtains the stress gradient and resistance change rate and combines them with the degradation characteristic coefficient to weighted analyze the initial response threshold, calls the temperature gradient data for amplitude correction, generates the dynamic response threshold, and transmits it to the risk assessment module. The risk assessment module extracts temperature gradient, stress gradient, and resistance change rate based on the state change data and calculates the response hysteresis. It performs amplitude comparison on the dynamic response threshold and analyzes the migration trend factor. It also makes risk judgments on temperature gradient, stress gradient, and resistance change rate and generates vehicle electrical performance monitoring results.

2. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 1, characterized in that: The state change data includes temperature change vector, stress change vector, and resistance change vector; the degradation characteristic coefficients include principal component coefficients, classification discrimination coefficients, and normalized difference coefficients; the dynamic response thresholds include stress correction thresholds, resistance correction thresholds, and temperature amplitude thresholds; and the vehicle electrical performance monitoring results include risk level index, response hysteresis index, and migration trend factor.

3. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 1, characterized in that, The data analysis module includes: The data stream receiving submodule collects the temperature, stress and resistance data of the fuse of the high-voltage connector, integrates them in chronological order, judges the acquisition error by comparing point by point, performs error elimination operation, and performs segment connection on the remaining sequence after elimination to generate the original monitoring sequence. The sequence smoothing submodule, based on the original monitoring sequence, calls the temperature, stress and resistance sequences, calculates the sequence volatility according to the difference values ​​of adjacent sampling points, and performs segment value replacement and segment mean reconstruction according to the volatility threshold to obtain a smoothed monitoring sequence. The state feature submodule, based on the smoothed monitoring sequence, calls the temperature, stress, and resistance sequences, calculates the temperature gradient and stress gradient, analyzes the resistance change rate based on the ratio of the resistance sequence before and after, and combines the temperature gradient, stress gradient, and resistance change rate to generate state change feature data.

4. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 3, characterized in that, The volatility threshold is determined by statistically analyzing the distribution of differences between adjacent sampling points in the sequence, calculating the absolute values ​​of the differences between the temperature, stress, and resistance sequences in chronological order, sorting all the absolute values ​​of the differences by numerical value to form a difference distribution sequence, and extracting the median value of the difference distribution sequence.

5. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 1, characterized in that, The feature extraction module includes: The gradient calculation submodule extracts temperature gradient, stress gradient value and resistance change rate based on the state change feature data, obtains the corresponding time series and performs difference comparison, filters out segments with stable change amplitude, organizes the gradient parameters at the same time position in a numerical accumulation manner, and generates gradient dataset. The normalization generation submodule calls the temperature gradient, stress gradient, and resistance change rate in the gradient dataset and performs scaling to calculate the normalized difference. It then filters the input parameters in the normalized difference by numerical comparison to generate the normalized input quantity of state change. The principal component classification submodule obtains the time series distribution of multiple parameters based on the normalized input of state change, calculates the covariance of the multi-parameter series to determine the trend correlation and selects the dominant trend component as the principal component of degradation trend, performs interval classification on the normalized input of state change, and generates degradation feature coefficients.

6. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 1, characterized in that, The threshold generation module includes: The gradient parameter submodule acquires the stress gradient and resistance change rate and performs corresponding time calibration. It uses numerical comparison to judge the data differences at multiple time points and eliminates abnormal jump points. Based on the data volume of the remaining segment, it performs sequence integration and synchronization to generate the stress-resistance joint sequence quantity. The degradation coefficient submodule calls the stress-resistance joint sequence quantity, performs weighting on the multi-time point data in the stress-resistance joint sequence quantity in combination with the degradation characteristic coefficient, removes the weighted results that exceed the amplitude benchmark value, and performs sequence accumulation based on the remaining weighted values ​​to obtain the weighted response basic result; The threshold correction submodule performs an amplitude comparison between the weighted response baseline result and the temperature gradient data at the same time point, calculates the deviation ratio between the amplitude value of the temperature gradient data and the preset standard amplitude, and superimposes it with the weighted response baseline result to generate a dynamic response threshold.

7. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 6, characterized in that, The amplitude reference value is determined by performing numerical distribution analysis on the data segment after weighting the degradation characteristic coefficients, obtaining the amplitude values ​​of all sampling points within the segment, sorting them, and calculating the median value of the sorted amplitude values.

8. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 1, characterized in that, The risk assessment module includes: The response hysteresis submodule extracts the time series of temperature gradient, stress gradient and resistance change rate based on the state change data, obtains the peak time points of multiple sequences and performs time difference calculation, uses the difference between peak time points to analyze the response difference, and generates the response hysteresis. The threshold comparison submodule calls the response hysteresis and the dynamic response threshold to perform amplitude segment comparison on the amplitude sequences of temperature gradient, stress gradient and resistance change rate, and performs amplitude difference calculation based on the difference between the amplitude sequences inside and outside the threshold segment to generate migration trend factor. The migration interval submodule calls the migration trend factor and the dynamic response threshold to perform interval division determination on temperature gradient, stress gradient and resistance change rate, and performs segment assignment determination and risk category classification between migration trend factor value and interval boundary to generate vehicle electrical performance monitoring results.

9. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 8, characterized in that, The system also includes: The condition monitoring module performs refined and graded processing based on the vehicle electrical performance monitoring results to obtain the fuse condition monitoring results, performs life decay trend analysis and feature clustering operation on the fuse condition monitoring results, and generates optimized vehicle electrical performance monitoring results. The optimized vehicle electrical performance monitoring results include the graded state index, life decay index, and feature clustering index.

10. The high-voltage connector fuse monitoring system and vehicle electrical performance monitoring system according to claim 9, characterized in that, The status monitoring module includes: The risk segmentation submodule, based on the vehicle electrical performance monitoring results, performs segment reading on the risk category, extracts the boundary difference between the segment position of the risk category and the risk classification benchmark value for classification judgment, calls the vehicle electrical operation index corresponding to the risk category as the classification mapping quantity, and generates fuse status monitoring results. The lifespan trend submodule calls the fuse status monitoring results, obtains the resistance change and temperature change, compares them with the risk level value, performs trend difference using the time series slope of the change sequence and the segment position of the risk level, analyzes the lifespan decay vector, and generates the lifespan decay trend. The feature clustering submodule calls the electrical behavior feature sequence of the life decay trend and the fuse status monitoring result and calculates the feature distance, extracts the amplitude range of the two and calculates the difference, and performs clustering in combination with the life decay trend value to generate optimized vehicle electrical performance monitoring results. The risk classification benchmark value is set by analyzing the distribution characteristics of electrical performance data, based on the long-term distribution characteristics of temperature, current, and voltage indicators, and by summing the mean of the long-term distribution characteristics and the standard deviation of three times the mean.

Images