Connector plug-in life prediction method and device based on electrical parameters and electronic equipment

Abstract

The application belongs to the technical field of reliability evaluation of electrical connectors, and discloses a connector plugging life prediction method and device based on electrical parameters and electronic equipment. The method comprises the following steps: collecting and preprocessing contact resistance time series data, and separating high-frequency measurement noise components through sliding window Hampel filtering; based on the irreversible mechanism of electrical contact degradation, a monotone non-decreasing irreversible long-term degradation component is extracted from the physical sequence by using the order-preserving regression method, an irreversible transient fluctuation component with mean reversion characteristics is obtained, and accurate decoupling of the two types of physical components is realized; a Wiener degradation model is constructed based on the irreversible component, parameters are identified, and the residual life distribution and confidence interval are calculated combined with the first arrival time theory; and dynamic iterative prediction is realized through local component updating and Bayesian parameter correction. The application can effectively eliminate the interference of reversible fluctuations and noise on parameter identification, and significantly improve the parameter accuracy of the Wiener model and the accuracy of early life prediction.

Description

Technical Field

[0001] This invention belongs to the field of electrical connector reliability assessment technology, specifically relating to a method, device, and electronic equipment for predicting connector mating and extraction life based on electrical parameters. Background Technology

[0002] Electrical connectors are fundamental core components in electronic and electrical systems, enabling power transmission and signal interaction. They are crucial for ensuring system circuit connectivity and signal integrity. As modern industrial equipment evolves towards higher integration, higher power density, and higher reliability, the number and application density of electrical connectors in systems continue to increase. A single complex piece of equipment often contains hundreds to thousands of pluggable electrical connectors. Contact failure in any one connector can trigger local circuit failures, signal transmission interruptions, or even system shutdowns, resulting in significant economic losses and safety risks. Therefore, the contact reliability of electrical connectors directly determines the operational stability and service safety of the entire electronic and electrical system. Accurately assessing and predicting the remaining mating life of connectors has significant engineering application value and economic significance for conducting condition-based maintenance, reducing system downtime risks, and preventing safety accidents.

[0003] During repeated mating and extraction, the contact interface of a pluggable electrical connector undergoes a series of complex physicochemical evolution processes, including progressive wear of the contact plating, transfer and exposure of the base material, generation and accumulation of wear debris, growth of the oxide film at the contact interface, and a continuous reduction in the effective conductive contact area. These processes collectively lead to an overall increasing degradation trend in contact resistance. When the contact resistance rises above the failure threshold specified in the product standard, the connector experiences contact failure. Mating / extraction life is a core indicator for measuring the reliability of electrical connectors. Currently, the industry mainly obtains life data through accelerated mating / extraction life tests in laboratories. However, this method is time-consuming, costly, and only provides batch-level statistical lifespans, failing to provide personalized remaining life predictions for individual connectors in service.

[0004] Existing technologies for predicting the mating and unmating life of electrical connectors are mainly based on methods that address random degradation processes.

[0005] Among various stochastic degradation process models, the Wiener stochastic degradation process model stands out due to its mature mathematical properties, high analytical performance, and clear physical meaning. It can simultaneously characterize both the long-term degradation trend and stochastic fluctuation characteristics, making it the most widely used core model with the highest compatibility with "connector mating and removal life prediction based on electrical parameters." This technology uses contact resistance as the core electrical parameter characterizing connector contact performance degradation. Its basic principle is as follows: with increasing mating and removal cycles, the connector contact plating continuously wears down, and oxide debris accumulates at the contact interface, resulting in an overall upward degradation trend in contact resistance with stochastic fluctuations. The Wiener process models the change in contact resistance with mating and removal cycles as a stochastic process containing drift and diffusion terms. The drift term characterizes the irreversible long-term degradation trend caused by mating and removal wear, corresponding to the actual lifespan decay rate of the connector. The diffusion term characterizes the stochastic oscillations caused by the redistribution of contact interface debris and changes in microscopic contact state during each mating and removal cycle, corresponding to the inherent stochasticity of the degradation process.

[0006] Through in-depth analysis of the physical processes at the contact interface during connector insertion and removal, and verification with a large amount of experimental data, the actually acquired contact resistance timing signal contains three types of fluctuation components with completely different physical properties and variation patterns: The first category is measurement noise components, caused by inherent random errors in the resistance acquisition system, on-site electromagnetic interference, mechanical jitter during insertion and removal, and transient contact bounce. These components are high-frequency random interferences without physical meaning, do not reflect the true degradation state of the contact interface, and have statistical characteristics of small amplitude, high frequency, and zero mean. They should be filtered out before modeling.

[0007] The second type is reversible transient physical fluctuation components, caused by physical processes such as the instantaneous accumulation and extrusion of wear debris at the contact interface, the elastic deformation and compaction of contact micro-protrusions, the instantaneous displacement of the contact point, and the dynamic distribution of the lubricating film during insertion and removal. These fluctuations reflect the true physical state of the contact interface and are not measurement noise. Their changes exhibit short-term reversibility and mean-recovery characteristics, meaning the contact resistance will fluctuate within a certain range and will not accumulate continuously with the number of insertions and removals. Therefore, they are not part of an irreversible degradation process and should not be included in the long-term degradation trend. The amplitude of this type of component is usually much larger than the measurement noise and is the main factor causing short-term fluctuations in contact resistance.

[0008] The third category is irreversible long-term degradation components, caused by irreversible physicochemical processes such as permanent wear of the contact plating, exposure and oxidation of the substrate material, permanent reduction of the contact conductive area, and continuous accumulation of oxide film at the contact interface. This type of component monotonically increases with the number of mating cycles and is the core degradation characteristic that truly reflects the decline in connector lifespan. It corresponds to the long-term degradation trend described by the drift term in the Wiener model and is also the core focus of lifespan prediction.

[0009] Existing traditional Wiener degradation modeling techniques do not differentiate or decouple the three types of components mentioned above. They typically only perform simple moving average or Kalman filtering on the original sequence, or even directly treat the original contact resistance sequence as the object of observation for the Wiener process. This approach has the following clear technical problems: First, traditional methods conflate reversible transient physical fluctuations with measurement noise, resulting in a fundamental contradiction in their processing. If strong filtering is used to smooth all fluctuations, the true physical state information of the contact interface will be lost, and even subtle early degradation characteristics may be masked, leading to the failure of early lifetime prediction. If fluctuations are not distinguished and are directly included in the diffusion term, the diffusion coefficient will be significantly overestimated, resulting in an unreasonable widening of the calculated lifetime confidence interval. This significantly reduces the reference value of the prediction results and fails to provide accurate basis for maintenance decisions.

[0010] Secondly, traditional methods cannot separate reversible transient components from irreversible degradation components, causing the drift coefficient identification results to be severely interfered with by reversible fluctuations. The drift coefficient is a core parameter determining the accuracy of lifetime prediction trends. When the reversible transient component shows a phased decrease, the model may misinterpret it as a decrease in the degradation rate or even a cessation of degradation, leading to an underestimation of the drift coefficient, an overly optimistic lifetime prediction result, and an inability to issue timely failure warnings, increasing the risk of system operation. Conversely, when the reversible transient component shows a phased rapid increase, the model may misinterpret it as an acceleration of the degradation rate, leading to an overestimation of the drift coefficient, an overly conservative lifetime prediction result, unnecessary premature maintenance, and a significant increase in operation and maintenance costs. Summary of the Invention

[0011] The purpose of this invention is to provide a connector mating and extraction life prediction method based on electrical parameters, and to provide a connector mating and extraction life prediction device and an electronic device based on electrical parameters, so as to solve the problems mentioned in the background art.

[0012] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0013] A connector mating and extraction life prediction method based on electrical parameters includes the following steps:

[0014] S1. Collect contact resistance data during the insertion and removal process of electrical connectors. After outlier removal and missing value completion, a standardized original contact resistance sequence is obtained. S2. The standardized original sequence is processed by using a sliding window Hampel filter to separate the high-frequency measurement noise components and obtain a physical sequence of contact resistance that only reflects the physical state of the contact interface. S3. Based on the irreversible mechanism of electrical contact degradation, the physical sequence is solved by order-preserving regression with monotonically non-decreasing constraints. The irreversible long-term degradation component is extracted and the residual is the reversible transient fluctuation component, thus achieving decoupling of the two types of physical components. S4. Using irreversible long-term degradation components as the modeling object, construct the Wiener degradation model and use maximum likelihood estimation to identify the drift coefficient and diffusion coefficient. S5. Based on the first arrival time theory, derive the lifetime distribution of the contact resistance when it reaches the failure threshold, and calculate the point estimate of the remaining lifetime and the corresponding confidence interval. S6. After the newly added plug-in data is accessed, the local components are decoupled and updated, and the model parameters are corrected by Bayesian method. The remaining lifetime is calculated iteratively to achieve dynamic prediction of the entire life cycle.

[0015] Furthermore, step S1 specifically includes: taking a single complete insertion / removal cycle as a unit, collecting multiple resistance sampling values ​​in the insertion stability segment of each cycle, taking the arithmetic mean as the contact resistance characteristic value of that insertion / removal; using the Grubbs criterion to remove measurement fault-type outliers, using linear interpolation to fill in a small number of missing points, and obtaining an equally spaced standardized original sequence.

[0016] Furthermore, step S2 specifically includes: setting a sliding window with a window width greater than the noise correlation length and less than the reversible fluctuation duration period; calculating the median and median absolute deviation for the data in each window, converting them into equivalent standard deviations, and using 3 times the equivalent standard deviation as a threshold, replacing points exceeding the threshold with the median of the window; finally obtaining the physical sequence and noise components, and verifying the zero mean, normality, and low autocorrelation of the noise.

[0017] Furthermore, in step S3, the irreversible mechanism of electrical contact degradation is as follows: contact coating wear, substrate oxidation, and reduction of effective contact area are all irreversible processes. Therefore, the irreversible degradation component is monotonically non-decreasing with the number of insertions and removals. The optimization objective of order-preserving regression is to minimize the sum of squared errors between the fitted value and the physical sequence, with the constraint condition being that the sequence is monotonically non-decreasing.

[0018] Furthermore, in step S3, the pool-side violator algorithm is used to solve the order-preserving regression model. Specifically, it includes: initializing each data point as an independent data block, detecting violations from left to right, merging adjacent data blocks that violate monotonic constraints and calculating the weighted average, backtracking after merging, until there are no violations in the entire sequence, and finally mapping to obtain the irreversible degenerate component sequence.

[0019] Furthermore, step S3 also includes verification of the decoupling results: verifying point by point that the irreversible degenerate component satisfies the monotonically non-decreasing constraint; verifying that the reversible transient component is a stationary sequence with mean-recovery properties using the augmented Dickey-Fuller unit root test; if both verifications pass, the decoupling is confirmed to be effective.

[0020] Furthermore, step S4 specifically includes: constructing a linear drift Wiener degradation model, the model expression of which is... ,in The initial resistance, The drift coefficient, The diffusion coefficient is... The motion is standard Brownian motion. Based on the incremental independent normal distribution characteristics, a log-likelihood function is constructed, and the derivative is used to obtain the analytical maximum likelihood estimates of the drift coefficient and diffusion coefficient.

[0021] Furthermore, step S5 specifically includes: setting a contact resistance failure threshold, deriving the probability density function and cumulative distribution function of total lifetime and remaining lifetime based on the inverse Gaussian distribution characteristics of the first arrival of the Wiener process; calculating the expected value of the remaining lifetime as a point estimate, and using the Delta method to calculate the two-sided confidence interval and the one-sided confidence lower limit at the set confidence level.

[0022] Furthermore, step S6 specifically includes: after each set number of newly added insertion / removal data, performing local order-preserving regression to update the components only in the tail region of the sequence; using historical parameters as the prior distribution, updating the model parameters using the Bayesian method in combination with the newly added data, recalculating the remaining lifetime and confidence interval, and repeating this process cyclically to achieve dynamic prediction.

[0023] Furthermore, step S5 also includes: based on the remaining life prediction results and multi-level early warning thresholds, classifying the data into four levels: normal, attention, early warning, and alarm, and outputting corresponding status early warning information.

[0024] This application also discloses an electronic device, including:

[0025] At least one processor; and

[0026] A memory communicatively connected to the at least one processor; wherein,

[0027] The memory stores instructions that can be executed by the at least one processor, which enables the at least one processor to perform the connector mating and unmating life prediction method based on electrical parameters of the present invention.

[0028] Beneficial Effects: Compared with existing traditional degradation modeling techniques, this invention has the following advantages: it achieves hierarchical and precise decoupling of multiple property components with clear physical mechanisms. This invention adopts a hierarchical processing architecture, sequentially separating measurement noise components, reversible transient physical fluctuation components, and irreversible long-term degradation components, clearly distinguishing fluctuation signals with different physical properties and solving the core problem of unclear confusion among various fluctuations in traditional techniques. Specifically, the noise separation method based on sliding window Hampel filtering has strong robustness and can effectively filter out measurement interference while completely preserving the true physical fluctuation information of the contact interface, avoiding the problem of losing physical information in traditional strong filtering methods. The order-preserving regression decoupling method based on the irreversible mechanism constraint of electrical contact, constrained by clear physical laws, requires no additional hyperparameters and can accurately extract the irreversible degradation trend that truly reflects lifetime decay from complex fluctuations, with stable and reliable decoupling results.

[0029] This invention fundamentally improves the accuracy of Wiener model parameter identification. It uses the decoupled irreversible degradation component as the sole modeling object of the Wiener model, eliminating the interference of reversible transient fluctuations on the estimation of drift and diffusion coefficients. The drift coefficient accurately reflects the long-term true degradation rate of the connector, avoiding estimation bias due to the phased rise and fall of reversible fluctuations, thus ensuring the accuracy of the lifetime prediction trend. The diffusion coefficient accurately characterizes the inherent randomness of the degradation process, avoiding unreasonable amplification due to the superposition of reversible fluctuations, thus ensuring the accuracy of the lifetime confidence interval. This improvement in parameter identification accuracy lays a solid foundation for accurate lifetime prediction. Attached Figure Description

[0030] Figure 1 This is a flowchart of the overall method for predicting connector mating and extraction life based on electrical parameters according to the present invention.

[0031] Figure 2 This is a flowchart of the contact resistance data acquisition and standardization preprocessing in an embodiment of the present invention;

[0032] Figure 3 This is a flowchart of the sliding window Hampel filter noise separation process in an embodiment of the present invention;

[0033] Figure 4 This is a flowchart of the decoupling process for the order-preserving regression components in an embodiment of the present invention;

[0034] Figure 5 This is a flowchart of the remaining lifespan calculation and early warning determination in an embodiment of the present invention. Detailed Implementation

[0035] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0036] This invention provides a connector mating and extraction life prediction method based on electrical parameters, such as... Figure 1 As shown, the steps include:

[0037] S1. Collect contact resistance data during the insertion and removal process of electrical connectors. After outlier removal and missing value completion, a standardized original contact resistance sequence is obtained. S2. The standardized original sequence is processed by using a sliding window Hampel filter to separate the high-frequency measurement noise components and obtain a physical sequence of contact resistance that only reflects the physical state of the contact interface. S3. Based on the irreversible mechanism of electrical contact degradation, the physical sequence is solved by order-preserving regression with monotonically non-decreasing constraints. The irreversible long-term degradation component is extracted and the residual is the reversible transient fluctuation component, thus achieving decoupling of the two types of physical components. S4. Using irreversible long-term degradation components as the modeling object, construct the Wiener degradation model and use maximum likelihood estimation to identify the drift coefficient and diffusion coefficient. S5. Based on the first arrival time theory, derive the lifetime distribution of the contact resistance when it reaches the failure threshold, and calculate the point estimate of the remaining lifetime and the corresponding confidence interval. S6. After the newly added plug-in data is accessed, the local components are decoupled and updated, and the model parameters are corrected by Bayesian method. The remaining lifetime is calculated iteratively to achieve dynamic prediction of the entire life cycle.

[0038] To more clearly illustrate the technical solution and implementation process of the present invention, the present invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0039] A connector mating and extraction lifetime prediction method based on electrical parameters is executed sequentially according to the following six steps: data acquisition and preprocessing, measurement noise separation, physical component decoupling, degradation model construction, lifetime calculation and output, and dynamic update iteration.

[0040] like Figure 2 As shown, step 1, connector contact resistance timing data acquisition and standardization preprocessing, is to obtain standardized, continuous, and equally spaced contact resistance degradation timing data to provide standardized basic data support for subsequent component decoupling and lifetime modeling.

[0041] The specific implementation process is as follows: For the test environment setup, depending on the target model of the electrical connector, a mating life test platform or an online monitoring system is selected based on the application scenario. The mating life test platform is equipped with a precision servo mating mechanism, which can set and stably control test parameters such as mating rate, insertion depth, alignment deviation, and contact normal pressure. The number of mating cycles is recorded throughout the test to ensure accurate and unbiased counting. The contact resistance acquisition unit uses a four-wire milliohm measuring device to eliminate interference from lead resistance and contact voltage drop, with a measurement resolution of no less than 1 microohm, ensuring measurement accuracy and long-term stability.

[0042] Synchronous data acquisition is initiated by starting the insertion / removal test or connecting to the online monitoring data stream. A single complete insertion / removal cycle is used as the basic time unit, defined as a complete action encompassing three phases: insertion, engagement, and removal. During the period of insertion and stable contact in each insertion / removal cycle, a predetermined number of contact resistance sampling points are continuously collected. The sampling frequency is set according to the test conditions and hardware performance to ensure complete coverage of the stable contact phase. The sampling start time is synchronized with the engagement signal in the hardware to eliminate time offset errors.

[0043] Single-cycle feature value extraction involves collecting multi-point sampling data within each insertion / removal cycle, discarding transient transitional data, and extracting the middle segment data where the contact state is stable after insertion. This segment is then used as the effective sampling segment. The arithmetic mean of all sampled values ​​within this effective sampling segment is calculated, and this average is used as the contact resistance feature value corresponding to this insertion / removal cycle. The length of the effective sampling segment is not less than 50% of the total sampling time of the single cycle to ensure that the feature value can stably represent the contact state after insertion / removal.

[0044] Outlier removal across the entire sequence involves using the Grubbs criterion to detect outliers in the original sequence comprised of contact resistance characteristic values ​​corresponding to all insertion and removal cycles. A standard significance level of 0.05 is set. The sample mean and standard deviation of the sequence are calculated. For each data point, the Grubbs statistic is calculated and compared to the critical value for the corresponding sample size. Data points with a statistic greater than the critical value are considered outliers and removed. Outlier detection only applies to outliers caused by measurement malfunctions, transient bounces, etc., and does not cover normal physical fluctuations.

[0045] Missing data completion is performed using linear interpolation to fill in missing data points resulting from outlier removal, as well as a small number of missing data points caused by temporary equipment interference or data transmission interruptions during testing or monitoring. For intervals with no more than three consecutive missing points, the linear interpolation result of the two valid data points before and after the missing interval is used for filling. For intervals with more than three consecutive missing points, they are marked as data faults and segmented for subsequent modeling. After completion, the continuity and equal intervals of the time series are maintained, and each interpolation cycle corresponds to a unique feature value.

[0046] The output is a standardized sequence. After the above processing, a standardized original contact resistance sequence with the number of insertions / removals (n) as the independent variable is obtained. Where n is a positive integer, and its value range is 1. N represents the total number of insertion / removal cycles completed so far. The sequence indexes correspond one-to-one with the insertion / removal counts, with no misalignment or offset.

[0047] like Figure 3 As shown, step 2, measurement noise component separation based on sliding window Hampel filtering, is a step that separates high-frequency random noise introduced by the measurement system and environmental interference from the standardized original sequence, and obtains a physical sequence of contact resistance that only reflects the true physical state of the contact interface, thus avoiding interference from measurement noise on subsequent component decoupling and model parameter identification.

[0048] The specific implementation process is as follows: The sliding window parameters are set based on the contact resistance sampling interval, insertion / removal rate, and typical duration of reversible transient fluctuations. The window width *w* is determined by two constraints: first, it must be greater than the correlation length of the measurement noise to ensure effective statistical analysis of the noise's characteristics; second, it must be less than the average duration of the reversible transient fluctuations to ensure that the filtering process does not smooth out the true physical fluctuation information. For typical insertion / removal test scenarios, the window width *w* is set to 7, meaning each window contains 7 consecutive contact resistance characteristic values.

[0049] The sliding window traversal process uses a sliding window of width w, starting from the beginning of the sequence and sequentially processing the original contact resistance sequence along the insertion / removal count axis. Perform traversal processing with a sliding step of 1. For each subsequence within a window, perform the following operations: calculate the median M of all data within the window (the median is highly robust and unaffected by extreme noise points); calculate the median absolute deviation (MAD) of all data within the window using the following formula: ;

[0050] in For the i-th data point within the window, To obtain the median function, the absolute deviation of the median is converted to the equivalent standard deviation using a standard transformation factor of 1.4826. ,Right now This conversion factor is applicable to normally distributed data and can guarantee the statistical consistency between the equivalent standard deviation and the sample standard deviation.

[0051] Set a fixed noise threshold coefficient k, with a value of 3, and divide the interval As the normal data range; for each data point within the window, if its value exceeds the above normal data range, it is determined to be a measurement noise point, and the original value of the point is replaced with the window's median value M; if the value is within the normal range, the original value is retained unchanged.

[0052] For boundary processing, a symmetrical expansion method is used to construct windows for boundary data points at the beginning and end of the sequence. That is, at the beginning of the sequence, subsequent data points are mirrored and copied before the first data point to make up the window length; at the end of the sequence, preceding data points are mirrored and copied after the last data point to make up the window length, ensuring that the filtering effect of the first and last data points is consistent with that of the middle data.

[0053] After extracting the physical sequence and noise components, and performing sliding window filtering on the entire sequence, the filtered physical sequence of contact resistance is obtained. The difference between the original sequence and the physical sequence is defined as the measurement noise component. ,Right now: ; Noise separation effect verification, based on the obtained measured noise components. Three statistical property checks were performed: zero mean check by calculating the mean of the noise sequence and verifying that its absolute value is less than 1% of the amplitude of the physical sequence, which approximately satisfies the zero mean property; normality check by using the Shapiro-Wilk test to verify that the noise sequence approximately follows a normal distribution, with a significance level of 0.05; and low autocorrelation check by calculating the first-order autocorrelation coefficient of the noise sequence and verifying that its absolute value is less than 0.1, indicating weak autocorrelation of the sequence, which is consistent with the characteristics of white noise.

[0054] If all three checks pass, the noise separation effect is considered satisfactory; if the checks fail, adjust the window width w or the threshold coefficient k, and re-execute the filtering process until the check requirements are met.

[0055] like Figure 4 As shown, step 3, decoupling the irreversible degradation component and the reversible transient component based on order-preserving regression, is the core step of this invention. Its function is to separate the monotonically non-decreasing irreversible long-term degradation component and the reversible transient fluctuation component with mean-recovery characteristics from the physical sequence of contact resistance based on the irreversible physical mechanism constraint of electrical contact degradation, thereby achieving precise decoupling of two types of fluctuation signals with different physical properties.

[0056] The specific implementation process is as follows: With the decoupling mechanism established, based on the fundamental physical principles of electrical contact degradation, the wear of the contact plating, exposure and oxidation of the substrate material, continuous growth of the oxide film at the contact interface, and permanent reduction of the contact conductive area during connector insertion and removal are all thermodynamically irreversible processes. They accumulate only with increasing insertion and removal cycles and do not recover spontaneously. Therefore, the irreversible degradation component characterizing the true degradation state... It must satisfy the constraint that it is monotonically non-decreasing, that is, for any positive integer... , always: ; This constraint is the core physical basis for distinguishing between irreversible degradation and reversible fluctuations, and it is also the core mechanism basis of the decoupling method of this invention.

[0057] The order-preserving regression optimization model is constructed using the physical sequence of contact resistance obtained in step 2. Using the observed values ​​as input, the number of insertions / removals (n) as the independent variable, and monotonically non-decreasing as the constraint, an order-preserving regression model is constructed. The optimization objective of the order-preserving regression is to minimize the sum of squared errors between the fitted values ​​and the observed values, and its mathematical expression is: ; ; in, Let be the sequence of irreversible degenerate components to be solved. The physical meaning of this optimization model is to find the monotonic sequence with the smallest deviation from the observed physical sequence, under the physical constraint of irreversible degeneracy being monotonically non-decreasing, as the true irreversible degenerate component.

[0058] The algorithm employs the Pool Adjacent Violators Algorithm (PAVA) to solve the above order-preserving regression model, obtaining the optimal fitted sequence that satisfies the monotonically non-decreasing constraint. This refers to the irreversible long-term degradation component. The algorithm merges adjacent data blocks that violate monotonic constraints, iteratively solves for the weighted average of each data block, and finally obtains the globally optimal monotonic fitting result. It boasts high computational efficiency, stable and reliable results, and requires no additional hyperparameters. The specific execution flow of the PAVA algorithm is as follows: Initialization: Each data point is treated as an independent data block, with each block's value representing the corresponding data point's value. Each block has a weight of 1, and all data blocks are arranged in order of insertion / removal count. Violation detection: All data blocks are traversed from left to right. If two adjacent data blocks violate the monotonic non-decreasing constraint (i.e., the value of the previous block is greater than the value of the next block), a violation is identified, and the violation position is recorded. Violation merging: Two adjacent data blocks violating the constraint are merged into a new data block. The value of the new block is the weighted average of the two blocks, calculated using the formula: ; in , The values ​​are the values ​​of the two original data blocks. , The weights of the two original data blocks are given by the weights of the two blocks; the weight of the new block is the sum of the weights of the two blocks, i.e. ; Backtracking check: After merging, backtrack one data block and check if there are still violations in the newly merged block compared to the previous block. If so, continue the merging operation until there are no violations at the current position. Termination condition: Repeat the violation detection and merging operation until all data blocks have been traversed from left to right without any violations, at which point the algorithm terminates. Result mapping: Map the value of each data block back to the corresponding position in the original sequence, that is, all data points within the same block take the value of that block, finally obtaining the complete irreversible degenerate component sequence. .

[0059] Reversible transient component extraction involves subtracting the irreversible degenerate component from the physical sequence of contact resistance to obtain the residual sequence, which is the reversible transient fluctuation component. The calculation formula is: ; The reversible transient fluctuation component reflects the resistance fluctuation caused by short-term reversible physical processes at the contact interface. It fluctuates around zero and has mean-recovery characteristics.

[0060] The decoupling results are validated by verifying the characteristics of the two decoupled components to ensure that the decoupling results conform to the physical mechanism. The irreversible degenerate component is verified as a point-by-point traversal sequence. Verify that the value of each subsequent time step is not less than the value of the previous time step, strictly satisfying the monotonically non-decreasing property. If there is a violation point, the solution is re-executed. The reversible transient component verification is performed using the augmented Dickey-Fuller (ADF) unit root test on the sequence. A stationarity test is performed with a significance level of 0.05. If the test result rejects the null hypothesis that a unit root exists, the sequence is confirmed to be a stationary sequence with mean-recovery properties, which conforms to the physical characteristics of reversible transient fluctuations.

[0061] If both checks pass, the decoupling result is confirmed to be valid; if the checks fail, check the quality of the input data, or adjust the preprocessing parameters and re-execute the decoupling calculation.

[0062] Step 4: Construction and parameter identification of Wiener degradation model based on the decoupled irreversible components. The function of this step is to construct a Wiener stochastic degradation process model with the decoupled irreversible degradation components as the modeling object, identify the key parameters of the model, eliminate the interference of reversible transient fluctuations on the model parameter estimation, and ensure that the parameters can truly reflect the long-term degradation law of the connector.

[0063] The specific implementation process is as follows: Wiener degradation model construction, including irreversible degradation components As an observation of the Wiener stochastic process, a linear drift Wiener degradation model is constructed. For any number of insertion / removal cycles n, the degradation of the contact resistance can be expressed as: ; in, The initial contact resistance value, i.e. The reference value for contact resistance at that time; It is the drift coefficient, with the unit being resistance per insertion and removal number of times. It characterizes the long-term degradation rate of contact resistance per unit number of insertion and removal times, reflecting the speed of irreversible degradation. The diffusion coefficient characterizes the inherent random fluctuation amplitude of the degradation process; It is a standard Brownian motion that satisfies the independent increment property, meaning that the increments within any non-overlapping time intervals are independent of each other.

[0064] The incremental distribution is derived based on the independent incremental characteristics of the Wiener process, where the increments of the irreversible degenerate components follow a normal distribution. Let the interval between adjacent insertions / removals be... Then the increment Follow the mean variance is The normal distribution is: ; Furthermore, each adjacent increment is independent of the others.

[0065] The log-likelihood function is constructed based on the aforementioned incremental independent normal distribution property, combined with the complete irreversible degenerate component sequence, to build the drift coefficient. and diffusion coefficient The log-likelihood function of . For an irreversible degenerate sequence of length N, there are a total of For each incremental sample, the log-likelihood function is expressed as: ; Parameter maximum likelihood estimation involves using the maximum likelihood estimation method to solve for the maximum value of the log-likelihood function, thereby obtaining the drift coefficients. and diffusion coefficient The optimal estimate. The log-likelihood function with respect to... and Taking the partial derivatives and setting them to zero, we can derive the analytical estimate of the parameters: ; ; in This is the maximum likelihood estimate of the drift coefficient. This is the maximum likelihood estimate of the variance of the diffusion coefficient.

[0066] Parameter rationality verification involves substituting the identified parameters into the Wiener model and generating multiple simulated degradation trajectories using the Monte Carlo method. The statistical characteristics of the simulated trajectories, including mean trend and variance growth rate, are calculated and compared with the evolutionary statistical characteristics of actual irreversible degradation components to verify the rationality of the parameters. If the deviation exceeds a set threshold, the accuracy of the decoupling results is checked, and parameter identification is re-executed.

[0067] like Figure 5 As shown, step 5, the remaining service life calculation and confidence interval estimation based on the first arrival time theory, is a step that calculates the remaining insertion and removal life of the connector and its probability distribution based on the identified Wiener model parameters and the preset failure threshold, and outputs the life interval under different confidence levels, providing a quantitative basis for reliability assessment and maintenance decisions.

[0068] The specific implementation process is as follows: Failure threshold setting: Based on the product technical specifications of the target connector, relevant industry standards, or failure criteria for engineering applications, a contact resistance failure threshold is set. When the contact resistance first reaches or exceeds this threshold, the connector is considered to have experienced contact failure, and the corresponding number of mating and unmating cycles is the total lifespan. The failure threshold must be set strictly in accordance with the standard requirements of the corresponding product and must not be changed arbitrarily.

[0069] First-arrival time distribution derivation: Based on the first-arrival time theory of the Wiener process, the derivation is performed to show when the contact resistance first reaches the failure threshold. The probability density function and cumulative distribution function at time T (i.e., total lifetime T) are given. For a linear drift Wiener process, the first arrival time follows an inverse Gaussian distribution, and the probability density function of the total lifetime T is: ; The cumulative distribution function of total lifetime is: ; in This is the cumulative distribution function of the standard normal distribution.

[0070] Derivation of remaining service life distribution, for the current number of insertions / removals N and the current degradation amount as... The remaining useful life (RUL) is defined as the number of insertions and removals from the current moment until the first failure. According to the Markov property of the Wiener process, the remaining lifetime also follows an inverse Gaussian distribution, and its probability distribution can be obtained by replacing the initial value with the current degradation amount and the time starting point with the current moment.

[0071] Point estimation of remaining lifetime: The expected value of the remaining lifetime is calculated and used as the point estimate of the remaining insertion and removal lifetime. The calculation formula is as follows: ; Confidence interval estimation, based on the identified model parameters and the asymptotic variance of the parameter estimates, uses the delta method to calculate the two-sided confidence interval and one-sided lower confidence limit of the remaining useful life at a set confidence level, quantifying the uncertainty of the life prediction. The default confidence levels are 90% and 95%, commonly used in engineering, but can be adjusted according to actual needs.

[0072] The warning level is determined based on the predicted remaining lifespan and pre-set multi-level warning thresholds. These thresholds are set according to the maintenance cycle and operational strategy, typically categorized into four levels: Normal, Attention, Warning, and Alarm. Each level corresponds to a different remaining lifespan range, outputting relevant status warning information to guide maintenance decisions.

[0073] Step 6, Dynamic Model Update and Iterative Prediction in Online Monitoring Scenarios. The function of this step is to enable rapid model updates and lifespan iterative predictions after new data is input during the continuous service of the connector or the ongoing testing, so as to meet the real-time requirements of online monitoring and ensure that the prediction accuracy continues to improve as the amount of data increases.

[0074] The specific implementation process is as follows: When new data is input, after collecting contact resistance data for M new insertion / removal cycles, the new data is preprocessed according to the method in step 1, including feature value extraction, outlier removal, and missing value completion, to obtain a new standardized contact resistance sequence. This new sequence is then appended to the end of the original sequence to form an updated original sequence. The amount of new data M is set according to the update cycle, typically ranging from 10 to 50 insertion / removal cycles, balancing real-time updates and computational efficiency.

[0075] Local component decoupling updates employ a sliding window-based order-preserving regression update strategy. This strategy only performs local order-preserving regression on the tail region of the sequence corresponding to newly added data, updating both irreversible degenerate and reversible transient component sequences. This eliminates the need to recalculate the entire sequence, thus improving computational efficiency. The window size for local updates is set to the length of the newly added data plus the length of the original tail data, ensuring consistency between the local decoupling results and the full decoupling results.

[0076] The model parameters are updated using Bayesian methods. Using the previously identified model parameters as the prior distribution and incorporating newly added irreversible degenerate component data, the updated drift and diffusion coefficients are obtained from the posterior distributions, enabling dynamic correction of the model parameters. The prior distribution of the drift coefficients is set to a normal distribution, while the prior distribution of the diffusion coefficients is set to an inverse gamma distribution. The selection of conjugate priors ensures the existence of analytical solutions in the update process, resulting in high computational efficiency.

[0077] The remaining lifetime is iteratively calculated. Based on the updated model parameters and the current degradation state, the remaining lifetime and confidence interval are recalculated according to the method in step 5, and the prediction results and warning status are refreshed.

[0078] The process is repeated in a loop as the insertion and removal process continues, executing the above data access, component update, parameter update and lifespan calculation processes repeatedly to achieve dynamic lifespan prediction and status monitoring of the connector throughout its entire lifecycle.

[0079] This application also discloses a connector mating and extraction life prediction device based on electrical parameters, comprising:

[0080] The data acquisition and access module is used to read the raw contact resistance sampling data transmitted by the test bench / monitoring equipment and to synchronize and bind the number of insertions and removals and the sampling timestamp.

[0081] The timing preprocessing module is used to perform Grubbs outlier removal, linear missing value completion, and output a standardized raw contact resistance sequence.

[0082] The noise separation module is used to run the sliding window Hampel filtering algorithm to separate the measurement noise component from the physical sequence of contact resistance;

[0083] The component decoupling module is used to perform PAVA order-preserving regression based on monotonic constraints, and to separate irreversible degenerate components and reversible transient fluctuation components.

[0084] The model identification module is used to construct the Wiener model with irreversible components and solve for the drift coefficient and diffusion coefficient through maximum likelihood estimation.

[0085] The lifetime calculation module is used to calculate the remaining lifetime point estimate, each confidence interval, and graded early warning based on the first arrival time inverse Gaussian distribution.

[0086] The dynamic update module is used to add local update components of new data, correct model parameters using Bayesian methods, and iteratively refresh lifetime prediction results.

[0087] The parameter configuration module is used to configure global algorithm parameters such as window length, significance level, failure threshold, confidence level, and update step size.

[0088] The results storage module is used to persistently save the original time series, each layer component, model parameters, and lifetime prediction reports.

[0089] This application also discloses an electronic device, including: a processor, a memory, and a computer program stored in the memory and executable by the processor. When the processor executes the computer program, it implements any of the aforementioned connector insertion and removal life prediction methods based on electrical parameters.

[0090] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A connector mating and extraction life prediction method based on electrical parameters, characterized in that, Includes the following steps: S1. Collect contact resistance data during the insertion and removal process of electrical connectors. After outlier removal and missing value completion, a standardized original contact resistance sequence is obtained. S2. The standardized original sequence is processed by a sliding window Hampel filter to separate the high-frequency measurement noise components and obtain a physical sequence of contact resistance that only reflects the physical state of the contact interface. S3. Based on the irreversible mechanism of electrical contact degradation, the physical sequence is solved by order-preserving regression with monotonically non-decreasing constraints. The irreversible long-term degradation component is extracted and the residual is the reversible transient fluctuation component, thus achieving decoupling of the two types of physical components. S4. Using irreversible long-term degradation components as the modeling object, construct the Wiener degradation model and use maximum likelihood estimation to identify the drift coefficient and diffusion coefficient; S5. Based on the first arrival time theory, derive the lifetime distribution of contact resistance when it reaches the failure threshold, and calculate the point estimate of the remaining lifetime and the corresponding confidence interval. S6. After the newly added plug-in / plug-out data is accessed, the local components are decoupled and updated, and the model parameters are corrected using Bayesian methods. The remaining lifetime is calculated iteratively to achieve dynamic prediction of the entire life cycle.

2. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S1 includes: Taking a single complete insertion / removal cycle as a unit, multiple resistance sampling values ​​are collected during the stable insertion segment of each cycle, and the arithmetic mean is taken as the contact resistance characteristic value of that insertion / removal. The Grubbs criterion was used to remove outliers caused by measurement failures, and linear interpolation was used to fill in a small number of missing points to obtain a standardized original sequence with equal intervals.

3. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S2 includes: setting a sliding window with a window width greater than the noise correlation length and less than the reversible fluctuation duration period; calculating the median and median absolute deviation for the data in each window, converting them into equivalent standard deviations, and using 3 times the equivalent standard deviation as a threshold, replacing points exceeding the threshold with the median of the window; finally obtaining the physical sequence and noise components, and verifying the zero mean, normality, and low autocorrelation of the noise.

4. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S3 includes: The irreversible mechanism of electrical contact degradation is as follows: contact coating wear, substrate oxidation and reduction of effective contact area are all irreversible processes. Therefore, the irreversible degradation component is monotonically non-decreasing with the number of insertions and removals. The optimization objective of order-preserving regression is to minimize the sum of squared errors between the fitted value and the physical sequence, and the constraint condition is that the sequence is monotonically non-decreasing.

5. The connector mating and extraction life prediction method based on electrical parameters according to claim 4, characterized in that, S3 includes: using the pool-side violator algorithm to solve the order-preserving regression model, specifically including: initializing each data point as an independent data block, detecting violations from left to right, merging adjacent data blocks that violate monotonic constraints and calculating the weighted average, backtracking after merging, until there are no violations in the entire sequence, and finally mapping to obtain the irreversible degenerate component sequence.

6. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S3 includes verification of decoupling results: point-by-point verification that irreversible degenerate components satisfy monotonically non-decreasing constraints; augmented Dickey-Fuller unit root test to verify that reversible transient components are stationary sequences with mean-recovery properties; if both verifications pass, the decoupling is confirmed to be effective.

7. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S4 specifically includes: constructing a linear drift Wiener degradation model, the model expression of which is... ,in The initial resistance, The drift coefficient, Where is the diffusion coefficient. The motion is standard Brownian motion. Based on the incremental independent normal distribution characteristics, a log-likelihood function is constructed, and the derivative is used to obtain the analytical maximum likelihood estimates of the drift coefficient and diffusion coefficient.

8. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S5 specifically includes: setting a contact resistance failure threshold, deriving the probability density function and cumulative distribution function of total lifetime and remaining lifetime based on the inverse Gaussian distribution characteristics of the first arrival of the Wiener process; calculating the expected value of the remaining lifetime as a point estimate, and using the Delta method to calculate the two-sided confidence interval and the one-sided confidence lower limit at the set confidence level.

9. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S6 specifically includes: after each set number of newly added insertion / removal data, performing local order-preserving regression to update the components only in the tail region of the sequence; using historical parameters as the prior distribution, updating the model parameters using the Bayesian method in combination with the newly added data, recalculating the remaining lifetime and confidence interval, and performing this iteratively to achieve dynamic prediction.

10. The connector mating and extraction life prediction method based on electrical parameters according to claim 1, characterized in that, S5 also includes classifying the remaining lifespan prediction results and multi-level warning thresholds into four levels: normal, attention, warning, and alarm, and outputting corresponding status warning information.

Images