An underwater vehicle control system competing failure reliability evaluation method

Abstract

The application discloses a kind of underwater vehicle control system competition failure reliability evaluation method, belong to underwater vehicle reliability evaluation technical field, for underwater vehicle control system competition failure reliability evaluation, the multiple-source monitoring of underwater vehicle control system is preprocessed, and the reliability model of underwater vehicle control system is constructed;Non-time-lag parameter and time-lag related parameter are constructed, and the estimated value of non-time-lag parameter and time-lag related parameter is calculated using least square method, statistical inference and time-lag structure identification method;Introduce alternate iteration optimization mechanism to carry out parameter estimation, and the reliability of underwater vehicle control system is calculated.The application realizes the accurate modeling and reliability evaluation of the competition failure degradation process and impact coupling effect of underwater vehicle control system by constructing dynamic self-adaptive feature extraction model and time-lag related degradation theory, and using parameter estimation and alternate iteration optimization strategy.

Description

Technical Field

[0001] This invention discloses a method for assessing the reliability of underwater vehicle control system competition failures, belonging to the field of underwater vehicle reliability assessment technology. Background Technology

[0002] As core equipment for deep-sea exploration, seabed resource development, and long-term ocean observation, the control system of an underwater vehicle is a key unit for power output and attitude adjustment, directly determining the vehicle's motion performance, energy efficiency, and mission reliability. During long-term service, the control system must operate continuously in extreme marine environments such as high pressure, high humidity, strong corrosion, and complex fluid disturbances. Affected by both environmental factors and loads, internal components are prone to degradation phenomena such as insulation aging, contact fatigue, material corrosion, and signal drift, leading to gradual performance deterioration.

[0003] As service time increases, the degradation process of control systems exhibits significant time-varying and stochastic characteristics. Its degradation state is not only related to current operating conditions but also influenced by the hysteresis of historical operating conditions and environmental stresses, demonstrating a pronounced memory effect. Simultaneously, random and sudden events in the external environment, such as fluid shocks, voltage transients, and load fluctuations, can cause instantaneous impact damage to the system, leading to a sudden increase in the degradation rate or directly inducing failure. The coupling effect between the degradation process and external shocks results in complex competitive characteristics in the overall system failure, making it difficult for traditional models to accurately describe its reliability variation patterns.

[0004] Therefore, there is a need to establish a competitive failure reliability assessment method that can comprehensively characterize the degradation and evolution law, time delay correlation characteristics and external shock effects of the underwater vehicle control system. Summary of the Invention

[0005] The purpose of this invention is to provide a method for assessing the reliability of underwater vehicle control systems in the event of competing failures, in order to solve the problem that existing technologies cannot accurately assess the reliability of control systems that simultaneously exhibit time-delay-related degradation and sudden failures.

[0006] A method for assessing the reliability of competition failures in an underwater vehicle control system includes:

[0007] S1. Construct a multi-dimensional feature vector of the underwater vehicle control system, set a sliding time window based on the time series of the multi-dimensional feature vector, use dynamic adaptive principal component analysis to adaptively update the feature subspace of the sliding time window, construct a multi-scale functional index matrix based on the time scale of the sliding time window, and construct the system functional index vector using mutual information analysis.

[0008] S2. Define a comprehensive degradation characterization quantity based on the system function index vector, introduce time delay correlation to calculate the time delay influence term of degradation rate, establish a system degradation evolution model based on the time delay influence term, construct a time delay correlation degradation model of the control system under the influence of external shock based on the effect of shock on the degradation process, and construct a reliability model of the underwater vehicle control system.

[0009] S3. Construct non-time-delay parameters and time-delay-related parameters, set anomaly detection conditions to remove abrupt change intervals, construct a sample set for the steady degradation of the underwater control system, and calculate the initial estimation results of the non-time-delay parameters using the least squares method, statistical distribution fitting, and expectation-maximization algorithm; construct a residual sequence based on the initial estimation results of the non-time-delay parameters, identify the time-delay structure of the residual sequence, obtain the final time-delay order using the Akaike information criterion, solve the linearized model using the weighted least squares method, and obtain the initial estimation results of the time-delay-related parameters; introduce an alternating iterative optimization mechanism for parameter estimation, and calculate the reliability of the underwater vehicle control system based on the reliability model of the underwater vehicle control system and the parameter estimation results.

[0010] S1 includes S1.1, the multi-source monitoring signals of the underwater vehicle control system at time... The multidimensional feature vector is:

[0011] ;

[0012] In the formula, For thrust output, For rotational speed, For current, For voltage, For temperature, For vibration acceleration, It is the transpose symbol;

[0013] Normalize and smooth the feature components:

[0014] ;

[0015] In the formula, For the first A multidimensional feature vector of a monitoring signal After normalization and smoothing , for The mean, for standard deviation For monitoring signal index, The number of monitored signals.

[0016] S1 includes S1.2, which introduces a dynamic adaptive principal component analysis method. Based on the observed samples within the sliding time window, adaptive feature subspace updates are performed. Let the current principal component matrix be... System Functional Indicator Vector for:

[0017] ;

[0018] ;

[0019] Update the principal component matrix using small perturbation eigenvalue decomposition:

[0020] ;

[0021] In the formula, The incremental correction matrix driven by the new samples. The update step size of the principal component matrix;

[0022] S1 includes S1.3, defining a multi-scale window set. , For each time scale Calculate the corresponding dynamic characteristic statistics :

[0023] ;

[0024] In the formula, For time-scale indexing. , The mean, For the signal At a specific time scale Calculate the mean within the range. For variance, For the signal At a specific time scale Internal variance calculation For kurtosis, For the signal At a specific time scale Internal calculation of kurtosis;

[0025] Constructing a multi-scale functional index matrix :

[0026] .

[0027] S1 includes S1.4, introducing a degradation reference quantity to characterize the degradation level of the system. :

[0028] ;

[0029] In the formula, for The weights;

[0030] Based on mutual information analysis, calculate respectively and Mutual information values ​​between them:

[0031] ;

[0032] In the formula, For mutual information value, for and The joint probability distribution, , for and Marginal probability distribution, , and Obtained through kernel density estimation method;

[0033] Sort the mutual information values ​​in descending order to obtain an ordered sequence:

[0034] ;

[0035] In the formula, Indicates ranking Mutual information value;

[0036] Set a cumulative contribution rate threshold , Set system function indicator filtering conditions Select the minimum number of system functional indicators that meet the system functional indicator screening criteria. Before selection Each functional indicator is used as a key functional indicator to construct a system functional indicator vector. :

[0037] ;

[0038] In the formula, let This is the index of the system function indicator vector. , For the first A vector of system functional indicators.

[0039] S2 includes, S2.1, definition The comprehensive degradation characterization quantity in it is for:

[0040] ;

[0041] In the formula, The feature weight vector;

[0042] Introducing time-delay correlation, let the lag time be... , The time-delay effect of the current degradation rate of the system is:

[0043] ;

[0044] In the formula, For time-delay kernel functions, For state mapping function, For the historical degradation state of the continuous-time model, when the degradation rate of the control system is approximately linearly related to the historical degradation state, we take... When the degradation rate is nonlinearly amplified with the degradation level, the nonlinear form is taken. , This is a nonlinear state map in the form of a power function, used to quantify the nonlinear effect of historical degradation states on the current degradation rate. This represents the combined impact of historical degradation status on the current degradation rate.

[0045] S2 includes S2.2, which establishes a system degradation evolution model based on the time-delay influence term:

[0046] ;

[0047] In the formula, Let be the baseline degradation rate of the system in steady state. For variable degradation rate driven by external load, For external load, The diffusion coefficient is... For standard Brownian motion, This is the differential symbol.

[0048] S2 includes S2.3, a time-delay-dependent degradation model of the control system under the influence of external shocks, based on the effect of shocks on the degradation process:

[0049] ;

[0050] ;

[0051] ;

[0052] In the formula, and To replace the variable, For the index of impact, , To modify the kernel function, For the first The moment the impact occurred For the first The impact intensity of the second impact, For the impact mapping function, For the shock state coupling function, Let be the Dirac impulse function.

[0053] S2 includes S2.4, constructing a reliability model for the underwater vehicle control system:

[0054] ;

[0055] In the formula, For reliability, For degradation failure time, For sudden failure time, For the degradation process, The degradation failure threshold, The impact failure threshold, For probability, For any one.

[0056] S3 includes, S3.1, constructing non-time-delay parameters. , For impact arrival rate, Impact intensity distribution;

[0057] Set anomaly detection conditions:

[0058] ;

[0059] In the formula, For degenerate incremental sequences, This is the abnormal sensitivity coefficient. Within the sampling time interval The mean, Within the sampling time interval Standard deviation;

[0060] when If the anomaly detection conditions are met, The sampling interval at which the time point occurs is considered a sudden change interval affected by external shocks. After removing the sudden change intervals, a sample set of samples showing the steady degradation of the underwater control system is obtained. ,exist Simplified system degradation evolution model:

[0061] ;

[0062] In the formula, Δ is the sampling time interval. For random disturbance terms, The random disturbance term follows a normal distribution. Let be the variance of the random disturbance term, and be a variable that follows a distribution.

[0063] The simplified system degradation evolution model is solved using the least squares method to obtain... The estimated value , The estimated value and The estimated value ;

[0064] Based on the number of impacts and observation time length ,calculate The estimated value :

[0065] ;

[0066] based on Determined through statistical distribution fitting And solve using a regression model. If the impact intensity cannot be directly observed, the expectation-maximization algorithm is used to identify the statistical regularity of the impact intensity within the latent variable framework; finally, the initial estimates of the non-time-delay parameters are obtained by integrating the results. ;

[0067] S3 includes S3.2, constructing time-delay related parameters. , For the first First-order discrete lag coefficient, , For the memory modulation function related to the impact, Index for time delay order, , The time-delay kernel function describes the strength of degenerative memory over continuous time. The time-delay effect intensity coefficient, This is the memory decay coefficient;

[0068] based on Construct residual sequences :

[0069] ;

[0070] In the formula, The impact mapping function obtained during the parameter estimation stage The estimated value, In the current sampling interval The immediate cumulative damage caused by all external shock events occurring within the system to the amount of system degradation;

[0071] Perform time-delay structure identification on the residual sequence Calculate the sample autocorrelation function :

[0072] ;

[0073] In the formula, Total failure time, The sample mean of the residual sequence;

[0074] Set the threshold for low-order hysteresis terms , For residual sequence With lag order of The degree of direct correlation between time and time, satisfying , The calculation formula is:

[0075] ;

[0076] In the formula, The correlation coefficient;

[0077] Set the time delay order threshold and threshold of high-dimensional redundant lag terms ,when and At that time, the Akaike Information Criterion was introduced to automatically filter out lagging structures:

[0078] ;

[0079] In the formula, For the Akaike Information Criteria, The number of time-delay related parameters. For the corresponding lag order Down The likelihood function;

[0080] By comparing different Below Select to make The smallest lag order is taken as the final lag order;

[0081] like and It can be linearized and solved using weighted least squares:

[0082] ;

[0083] In the formula, Find the parameters that minimize the objective function. This represents the historical degradation state of the discrete-time model;

[0084] The initial estimates of the time-delay correlation parameters were finally obtained. .

[0085] S3 includes S3.3, which introduces an alternating iterative optimization mechanism to... , Initialize as the initial value, and then execute steps S3.3.1, S3.3.2, and S3.3.3 in a loop;

[0086] S3.3.1, Update the time delay parameters and fix them. Re-estimation using residual sequences ;

[0087] S3.3.2, Perform non-time-delay parameter correction, to Update known parameters ,right Perform a re-estimation;

[0088] S3.3.3 Perform convergence determination and set the convergence threshold. and If the relative error of parameter changes between two iterations satisfies:

[0089] ;

[0090] ;

[0091] The estimation result is determined to be convergent, and the following is obtained: and .

[0092] S3 includes, S3.4, and will and Substituting the data into the reliability model of the underwater vehicle control system, we obtain the reliability of the underwater vehicle control system.

[0093] Compared with existing technologies, this invention has the following advantages: This invention constructs a dynamic adaptive principal component analysis model and an online incremental feature extraction method, effectively extracting system state features, and establishes a time-delay-related degradation model for the degradation history lag effect, characterizing the delayed impact of historical degradation state and the instantaneous and cumulative effects of shocks; it proposes a parameter estimation and alternating iterative optimization strategy, which improves the stability of parameter estimation by separating the influence of historical degradation state, and achieves accurate identification of the nonlinear relationship between time delay and shock by using alternating iteration, thereby improving the model accuracy and reliability calculation accuracy. Attached Figure Description

[0094] Figure 1 This is a flowchart of the method of the present invention;

[0095] Figure 2 This is a diagram of the degradation process of an underwater vehicle control system based on time-delay correlation.

[0096] Figure 3 It is a simulation diagram of degraded data;

[0097] Figure 4 This is a graph showing the reliability results of the numerical simulation. Detailed Implementation

[0098] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0099] A method for assessing the reliability of competition failures in an underwater vehicle control system includes:

[0100] S1. Construct a multi-dimensional feature vector of the underwater vehicle control system, set a sliding time window based on the time series of the multi-dimensional feature vector, use dynamic adaptive principal component analysis to adaptively update the feature subspace of the sliding time window, construct a multi-scale functional index matrix based on the time scale of the sliding time window, and construct the system functional index vector using mutual information analysis.

[0101] S2. Define a comprehensive degradation characterization quantity based on the system function index vector, introduce time delay correlation to calculate the time delay influence term of degradation rate, establish a system degradation evolution model based on the time delay influence term, construct a time delay correlation degradation model of the control system under the influence of external shock based on the effect of shock on the degradation process, and construct a reliability model of the underwater vehicle control system.

[0102] S3. Construct non-time-delay parameters and time-delay-related parameters, set anomaly detection conditions to remove abrupt change intervals, construct a sample set for the steady degradation of the underwater control system, and calculate the initial estimation results of the non-time-delay parameters using the least squares method, statistical distribution fitting, and expectation-maximization algorithm; construct a residual sequence based on the initial estimation results of the non-time-delay parameters, identify the time-delay structure of the residual sequence, obtain the final time-delay order using the Akaike information criterion, solve the linearized model using the weighted least squares method, and obtain the initial estimation results of the time-delay-related parameters; introduce an alternating iterative optimization mechanism for parameter estimation, and calculate the reliability of the underwater vehicle control system based on the reliability model of the underwater vehicle control system and the parameter estimation results.

[0103] S1 includes S1.1, the multi-source monitoring signals of the underwater vehicle control system at time... The multidimensional feature vector is:

[0104] ;

[0105] In the formula, For thrust output, For rotational speed, For current, For voltage, For temperature, For vibration acceleration, It is the transpose symbol;

[0106] Normalize and smooth the feature components:

[0107] ;

[0108] In the formula, For the first A multidimensional feature vector of a monitoring signal After normalization and smoothing , for The mean, for standard deviation For monitoring signal index, The number of monitored signals.

[0109] S1 includes S1.2, which introduces a dynamic adaptive principal component analysis method. Based on the observed samples within the sliding time window, adaptive feature subspace updates are performed. Let the current principal component matrix be... System Functional Indicator Vector for:

[0110] ;

[0111] ;

[0112] Update the principal component matrix using small perturbation eigenvalue decomposition:

[0113] ;

[0114] In the formula, The incremental correction matrix driven by the new samples. The update step size of the principal component matrix;

[0115] S1 includes S1.3, defining a multi-scale window set. , For each time scale Calculate the corresponding dynamic characteristic statistics :

[0116] ;

[0117] In the formula, For time-scale indexing. , The mean, For the signal At a specific time scale Calculate the mean within the range. For variance, For the signal At a specific time scale Internal variance calculation For kurtosis, For the signal At a specific time scale Internal calculation of kurtosis;

[0118] Constructing a multi-scale functional index matrix :

[0119] .

[0120] S1 includes S1.4, introducing a degradation reference quantity to characterize the degradation level of the system. :

[0121] ;

[0122] In the formula, for The weights;

[0123] Based on mutual information analysis, calculate respectively and Mutual information values ​​between them:

[0124] ;

[0125] In the formula, For mutual information value, for and The joint probability distribution, , for and Marginal probability distribution, , and Obtained through kernel density estimation method;

[0126] Sort the mutual information values ​​in descending order to obtain an ordered sequence:

[0127] ;

[0128] In the formula, Indicates ranking Mutual information value;

[0129] Set a cumulative contribution rate threshold , Set system function indicator filtering conditions Select the minimum number of system functional indicators that meet the system functional indicator screening criteria. Before selection Each functional indicator is used as a key functional indicator to construct a system functional indicator vector. :

[0130] ;

[0131] In the formula, let This is the index of the system function indicator vector. , For the first A vector of system functional indicators.

[0132] S2 includes, S2.1, definition The comprehensive degradation characterization quantity in it is for:

[0133] ;

[0134] In the formula, The feature weight vector;

[0135] Introducing time-delay correlation, let the lag time be... , The time-delay effect of the current degradation rate of the system is:

[0136] ;

[0137] In the formula, For time-delay kernel functions, For state mapping function, For the historical degradation state of the continuous-time model, when the degradation rate of the control system is approximately linearly related to the historical degradation state, we take... When the degradation rate is nonlinearly amplified with the degradation level, the nonlinear form is taken. , This is a nonlinear state map in the form of a power function, used to quantify the nonlinear effect of historical degradation states on the current degradation rate. This represents the combined impact of historical degradation status on the current degradation rate.

[0138] S2 includes S2.2, which establishes a system degradation evolution model based on the time-delay influence term:

[0139] ;

[0140] In the formula, Let be the baseline degradation rate of the system in steady state. For variable degradation rate driven by external load, For external load, The diffusion coefficient is... For standard Brownian motion, This is the differential symbol.

[0141] S2 includes S2.3, a time-delay-dependent degradation model of the control system under the influence of external shocks, based on the effect of shocks on the degradation process:

[0142] ;

[0143] ;

[0144] ;

[0145] In the formula, and To replace the variable, For the index of impact, , To modify the kernel function, For the first The moment the impact occurred For the first The impact intensity of the second impact, For the impact mapping function, For the shock state coupling function, Let be the Dirac impulse function.

[0146] S2 includes S2.4, constructing a reliability model for the underwater vehicle control system:

[0147] ;

[0148] In the formula, For reliability, For degradation failure time, For sudden failure time, For the degradation process, The degradation failure threshold, The impact failure threshold, For probability, For any one.

[0149] S3 includes, S3.1, constructing non-time-delay parameters. , For impact arrival rate, Impact intensity distribution;

[0150] Set anomaly detection conditions:

[0151] ;

[0152] In the formula, For degenerate incremental sequences, This is the abnormal sensitivity coefficient. Within the sampling time interval The mean, Within the sampling time interval Standard deviation;

[0153] when If the anomaly detection conditions are met, The sampling interval at which the time point occurs is considered a sudden change interval affected by external shocks. After removing the sudden change intervals, a sample set of samples showing the steady degradation of the underwater control system is obtained. ,exist Simplified system degradation evolution model:

[0154] ;

[0155] In the formula, Δ is the sampling time interval. For random disturbance terms, The random disturbance term follows a normal distribution. Let be the variance of the random disturbance term, and be a variable that follows a distribution.

[0156] The simplified system degradation evolution model is solved using the least squares method to obtain... The estimated value , The estimated value and The estimated value ;

[0157] Based on the number of impacts and observation time length ,calculate The estimated value :

[0158] ;

[0159] based on Determined through statistical distribution fitting And solve using a regression model. If the impact intensity cannot be directly observed, the expectation-maximization algorithm is used to identify the statistical regularity of the impact intensity within the latent variable framework; finally, the initial estimates of the non-time-delay parameters are obtained by integrating the results. ;

[0160] S3 includes S3.2, constructing time-delay related parameters. , For the first First-order discrete lag coefficient, , For the memory modulation function related to the impact, Index for time delay order, , The time-delay kernel function describes the strength of degenerative memory over continuous time. The time-delay effect intensity coefficient, This is the memory decay coefficient;

[0161] based on Construct residual sequences :

[0162] ;

[0163] In the formula, The impact mapping function obtained during the parameter estimation stage The estimated value, In the current sampling interval The immediate cumulative damage caused by all external shock events occurring within the system to the amount of system degradation;

[0164] Perform time-delay structure identification on the residual sequence Calculate the sample autocorrelation function :

[0165] ;

[0166] In the formula, Total failure time, The sample mean of the residual sequence;

[0167] Set the threshold for low-order hysteresis terms , For residual sequence With lag order of The degree of direct correlation between time and time, satisfying , The calculation formula is:

[0168] ;

[0169] In the formula, The correlation coefficient;

[0170] Set the time delay order threshold and threshold of high-dimensional redundant lag terms ,when and At that time, the Akaike Information Criterion was introduced to automatically filter out lagging structures:

[0171] ;

[0172] In the formula, For the Akaike Information Criteria, The number of time-delay related parameters. For the corresponding lag order Down The likelihood function;

[0173] By comparing different Below Select to make The smallest lag order is taken as the final lag order;

[0174] like and It can be linearized and solved using weighted least squares:

[0175] ;

[0176] In the formula, Find the parameters that minimize the objective function. This represents the historical degradation state of the discrete-time model;

[0177] The initial estimates of the time-delay correlation parameters were finally obtained. .

[0178] S3 includes S3.3, which introduces an alternating iterative optimization mechanism to... , Initialize as the initial value, and then execute steps S3.3.1, S3.3.2, and S3.3.3 in a loop;

[0179] S3.3.1, Update the time delay parameters and fix them. Re-estimation using residual sequences ;

[0180] S3.3.2, Perform non-time-delay parameter correction, to Update known parameters ,right Perform a re-estimation;

[0181] S3.3.3 Perform convergence determination and set the convergence threshold. and If the relative error of parameter changes between two iterations satisfies:

[0182] ;

[0183] ;

[0184] The estimation result is determined to be convergent, and the following is obtained: and .

[0185] S3 includes, S3.4, and will and Substituting the data into the reliability model of the underwater vehicle control system, we obtain the reliability of the underwater vehicle control system.

[0186] When the sample data is in discrete form, the system degradation evolution model is transformed into a recursive discrete time-delay degradation model:

[0187] ;

[0188] In the formula, Let be a standard Gaussian random variable with a mean of 0 and a variance of 1, i.e. It is used to characterize the random perturbation uncertainty in the degradation process. The random disturbance intensity coefficient. Used to ensure the statistical consistency of discrete models under varying time scales;

[0189] The effects of shocks on degradation processes include the fact that in complex underwater environments, control systems are subject to sporadic external shocks such as sudden load changes, fluid disturbances, and voltage transients. These shocks can not only directly lead to transient failures but also alter the system's degradation evolution. Therefore, a shock process is introduced to describe the effect of sudden events on degradation. Assume that the arrival of external shocks follows a Poisson process with an impact arrival rate of [missing information]. , No. The time of the impact was Impact strength is The magnitude of the impact follows a distribution. .

[0190] The effects of impact on the degradation process mainly fall into two categories: direct failure and failure due to impact intensity. Exceeding the threshold When this happens, the system immediately fails, and its failure probability is expressed as:

[0191] ;

[0192] in, It is an exponential function. For sensitivity parameters, express ,when hour, This indicates that the impact was insufficient to cause direct failure; when hour, , representing the contribution of the impact to the probability of direct failure.

[0193] Degradation disturbances, including shocks, cause abrupt changes in the amount of degradation and have a lasting impact on the future degradation rate. The immediate jump effect of the shock is as follows:

[0194] ;

[0195] in, This is the mapping function from impact intensity to degradation increment. Simultaneously, to describe the impact's effect on the subsequent degradation rate, an impact correction kernel function is introduced. Its impact term is expressed as:

[0196] ;

[0197] In the formula, It is a cumulative amount representing the lingering, accelerating effect of historical shock events on the current rate of degradation.

[0198] The discretized form of the time-delay-related degradation model of the control system under external shocks is as follows:

[0199] ;

[0200] ;

[0201] ;

[0202] In the formula, and For the substitution variable.

[0203] The derivation of the reliability of underwater vehicle control systems involves considering time-delay correlations, noting that the impact effect not only manifests as instantaneous damage but also nonlinearly amplifies the subsequent degradation rate. To quantitatively describe the comprehensive impact of impact on degradation, the following indices are defined:

[0204] Instantaneous damage Characterizing the direct effect of the impact on the amount of degradation at the moment of occurrence:

[0205] ;

[0206] Accumulated impact , For a single impact in the future time window The cumulative promoting effect of internal degradation:

[0207] ;

[0208] In the formula, The duration of the impact effect is the window length.

[0209] Degradation Accelerator :

[0210] ;

[0211] in, For adjustment coefficients, This indicates that the impact accelerated the degradation.

[0212] By analyzing the degenerate increment sequence Anomaly detection identifies the time point of external shock occurrence, eliminates intervals containing abrupt changes, and obtains a sample set of smoothly degrading underwater control systems. The steps include:

[0213] The degenerate increment sequence is represented as:

[0214] ;

[0215] In the formula, For the amount of degradation in The increment of time, where Δ is the sampling time interval;

[0216] For degenerate increment sequences Anomaly detection is performed. When an external shock occurs, the degradation increment will show abnormal fluctuations that significantly deviate from its normal statistical range. The anomaly criterion is defined as follows:

[0217] ;

[0218] in, The anomaly sensitivity coefficient is a positive real number between 2 and 4. Degenerate increment sequence within the sampling time interval The mean, Degenerate increment sequence within the sampling time interval The standard deviation.

[0219] When the above criteria are met, the corresponding time will be... The moment determined as the time of external impact will be recorded. The sampling interval is considered as a sudden change interval affected by external shocks. After removing intervals containing sudden changes, a sample set of samples showing the steady degradation of the underwater control system is obtained. .

[0220] The simplified system degradation evolution model is solved using the least squares method to obtain... The estimated value , The estimated value and The estimated value The steps include:

[0221] by As the dependent variable, with and Using the variables as independent variables, a regression sample is constructed. The least squares criterion is used to minimize the sum of squared prediction errors for all observations in the sample set, yielding the parameter vector. The estimated value:

[0222] ;

[0223] Based on this, the variance estimate of the random disturbance term is calculated from the regression residuals. :

[0224] .

[0225] Initial estimation results of non-time-delay parameters The solution process is as follows:

[0226] The set of moments in which external shocks occurred was identified. Then, shock response samples are constructed based on the abrupt change in degradation magnitude before and after the impact:

[0227] ;

[0228] In the formula, For the first The amount of degradation mutation caused by the second shock. To represent the nth Sampling time before the occurrence of the external impact. To indicate the first Sampling time after the occurrence of the external impact;

[0229] Assuming impact strength Follows a parameterized probability distribution Based on shock response samples The distribution parameters are determined using maximum likelihood estimation, and its likelihood function is:

[0230] ;

[0231] The statistical distribution of impact intensity is obtained by maximizing the log-likelihood function. :

[0232] ;

[0233] In the formula, for The parameter vector, The result obtained by the maximum likelihood estimation method The estimated value;

[0234] To describe the impact of a single impact on the system's degradation, a mapping relationship between impact intensity and degradation mutation is established:

[0235] ;

[0236] in, This is the random error term;

[0237] When the mapping function takes the form of a linear function When this happens, the impact mapping function can be obtained by solving the least squares criterion. The estimation results are as follows:

[0238] ;

[0239] In the formula, To represent the scaling factor in the impact mapping function, used to quantify the degree of influence of unit impact intensity on the abrupt degradation of the system. Impact strength The expected value of the degradation mutation caused by the impact mapping function acting on the system;

[0240] When impact strength When it cannot be directly measured, it is treated as a latent variable, and a joint probability model is constructed:

[0241] ;

[0242] In the formula, the parameter set ;

[0243] The parameter estimation using the Expectation-Maximization (EM) algorithm includes the following basic process: the E-step involves taking a sample of the current parameter estimates. Under the condition of calculating latent variables Conditional expectation:

[0244] ;

[0245] In the formula, In the expectation maximization algorithm, the first step is to... During the next iteration, the parameter set The estimated value; the M-step includes updating the expected result obtained from the E-step. and To maximize the expected log-likelihood function, the E-step and M-step are performed iteratively until the parameters converge, yielding the desired log-likelihood function. and The estimated value.

[0246] The overall failure of a control system arises from a competition between two mechanisms. achieve Degradation failure at that time; Exceed Sudden failure.

[0247] The degradation failure time is:

[0248] ;

[0249] In the formula, This is the infimum operator;

[0250] The sudden failure time is:

[0251] ;

[0252] The total failure time of the system is:

[0253] ;

[0254] In the formula, To find the minimum value;

[0255] Based on the above results, a reliability model for the control system of underwater vehicles is constructed.

[0256] The following explanation, in conjunction with the accompanying drawings, will provide further details. Figure 1 As shown, based on multi-source monitoring data of the underwater vehicle control system, a dynamic adaptive principal component analysis method is used to adaptively update the feature subspace of the sliding time window. A multi-scale functional index matrix is ​​constructed based on the time scale of the multi-scale window. Correlation assessment of the indices at each time scale is performed based on mutual information analysis, forming a system functional index vector. A comprehensive degradation characterization quantity is defined based on the system functional index vector. A time-delayed influence term for calculating the degradation rate is introduced, and a system degradation evolution model is established based on the time-delayed influence term. A time-delayed correlation degradation model under the influence of external shocks is constructed based on the effect of shocks on the degradation process, and a reliability model of the underwater vehicle control system is also constructed. Non-time-delay parameters and time-delayed correlation parameters are constructed based on the experience... After anomaly detection and separation, stationary degradation data and shock event data are used to calculate initial estimates of non-time-delay parameters using least squares, statistical distribution fitting, and expectation-maximization algorithms. Residual sequences are constructed based on the original degradation increments, and the autocorrelation and partial autocorrelation functions of the residual sequences are analyzed using time-delay structure identification methods. For the linearized model, weighted least squares is used for solution; for the nonlinearized model, extended Kalman filtering combined with expectation-maximization is used to achieve likelihood optimization in the state space, obtaining initial estimates of time-delay related parameters. An alternating iterative optimization mechanism is introduced for parameter estimation. Based on the underwater vehicle control system reliability model and the parameter estimation results, the reliability of the underwater vehicle control system is calculated.

[0257] Figure 2This diagram illustrates the degradation and evolution trajectory of the underwater navigator control system during long-term service, with the horizontal axis representing time. The vertical axis represents the system degradation. Horizontal dashed line The threshold represents the degradation failure threshold of the system. When the degradation amount first reaches this threshold, the system experiences degradation failure. The solid curve represents the degradation process of the system under the influence of time-delay correlation. The dashed line with arrows represents the time-delay correlation path in the degradation process of the system, which is used to describe the influence of the degradation state at the previous moment on the degradation evolution at subsequent moments.

[0258] Figure 3 The image shows the simulation degradation data of the control system generated using the Monte Carlo method, with and without considering the effects of time-delay correlation terms (set). , , , , , The degradation data considering time lag correlation introduces the lagged effects of historical degradation states into the current degradation increment. As can be seen from the figure, the degradation curve considering time lag correlation is basically consistent with the result without considering time lag correlation in the overall trend, but it shows more obvious correlation and cumulative dependence characteristics in the local evolution process. It can reflect the continuous impact of historical degradation state on subsequent degradation evolution, thus more realistically depicting the memory effect in the actual degradation process of the system.

[0259] Figure 4 As shown, based on simulated degradation data, the method proposed in this invention is applied to calculate the change in system reliability over time. The calculation results of the proposed method are highly close to the Monte Carlo simulation results throughout the entire time range, especially during the main operating phase of the system. This verifies that, considering the time delay correlation, the proposed method has high accuracy and can replace the computationally expensive Monte Carlo simulation with a higher confidence level for the reliability assessment of this system.

[0260] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for assessing the reliability of a control system for underwater vehicles in the event of race failures, characterized in that, include: S1. Construct a multi-dimensional feature vector of the underwater vehicle control system, set a sliding time window based on the time series of the multi-dimensional feature vector, use dynamic adaptive principal component analysis to adaptively update the feature subspace of the sliding time window, construct a multi-scale functional index matrix based on the time scale of the sliding time window, and construct the system functional index vector using mutual information analysis. S2. Define a comprehensive degradation characterization quantity based on the system function index vector, introduce time delay correlation to calculate the time delay influence term of degradation rate, establish a system degradation evolution model based on the time delay influence term, construct a time delay correlation degradation model of the control system under the influence of external shock based on the effect of shock on the degradation process, and construct a reliability model of the underwater vehicle control system. S3. Construct non-time-delay parameters and time-delay related parameters, set anomaly detection conditions to remove abrupt intervals, construct a sample set of steady degradation of the underwater control system, and use the least squares method, statistical distribution fitting and expectation-maximization algorithm to calculate the initial estimation results of non-time-delay parameters. Based on the initial estimation results of non-time-delay parameters, a residual sequence is constructed. After identifying the time-delay structure of the residual sequence, the final time-delay order is obtained using the Akaike information criterion. The linearized model is solved using the weighted least squares method to obtain the initial estimation results of the time-delay related parameters. An alternating iterative optimization mechanism is introduced for parameter estimation. Based on the reliability model of the underwater vehicle control system and combined with the parameter estimation results, the reliability of the underwater vehicle control system is calculated. Constructing system function indicator vectors : ； In the formula, let This is the index of the system function indicator vector. , The minimum number of system functional indicators, For the first A vector of system functional indicators; S2 includes, S2.1, definition The comprehensive degradation characterization quantity in it is for: ； In the formula, The feature weight vector; Introducing time-delay correlation, let the lag time be... , The time-delay effect of the current degradation rate of the system is: ； In the formula, For time-delay kernel functions, For state mapping function, For the historical degradation state of the continuous-time model, when the degradation rate of the control system is approximately linearly related to the historical degradation state, we take... ; When the degradation rate is nonlinearly amplified with the degradation level, the nonlinear form is taken. , This is a nonlinear state map in the form of a power function, used to quantify the nonlinear effect of historical degradation states on the current degradation rate. This represents the combined impact of historical degradation status on the current degradation rate. S2 includes S2.2, which establishes a system degradation evolution model based on the time-delay influence term: ； In the formula, Let be the baseline degradation rate of the system in steady state. For variable degradation rate driven by external load, For external load, The diffusion coefficient is... For standard Brownian motion, This is the differential symbol.

2. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 1, characterized in that, S1 includes S1.1, the multi-source monitoring signals of the underwater vehicle control system at time... The multidimensional feature vector is: ； In the formula, For thrust output, For rotational speed, For current, For voltage, For temperature, For vibration acceleration, It is the transpose symbol; Normalize and smooth the feature components: ； In the formula, For the first A multidimensional feature vector of a monitoring signal After normalization and smoothing , for The mean, for standard deviation For monitoring signal index, The number of monitored signals.

3. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 2, characterized in that, S1 includes S1.2, which introduces a dynamic adaptive principal component analysis method. Based on the observed samples within the sliding time window, adaptive feature subspace updates are performed. Let the current principal component matrix be... System Functional Indicator Vector for: ； ； Update the principal component matrix using small perturbation eigenvalue decomposition: ； In the formula, The incremental correction matrix driven by the new samples. The update step size of the principal component matrix; S1 includes S1.3, defining a multi-scale window set. , For each time scale Calculate the corresponding dynamic characteristic statistics : ； In the formula, For time-scale indexing. , The mean, For the signal At a specific time scale Calculate the mean within the range. For variance, For the signal At a specific time scale Internal variance calculation For kurtosis, For the signal At a specific time scale Internal calculation of kurtosis; Constructing a multi-scale functional index matrix : 。 4. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 3, characterized in that, S1 includes S1.4, introducing a degradation reference quantity to characterize the degradation level of the system. : ； In the formula, for The weights; Based on mutual information analysis, calculate respectively and Mutual information values ​​between them: ； In the formula, For mutual information value, for and The joint probability distribution, , for and Marginal probability distribution, , and Obtained through kernel density estimation method; Sort the mutual information values ​​in descending order to obtain an ordered sequence: ； In the formula, Indicates ranking Mutual information value; Set a cumulative contribution rate threshold , Set system function indicator filtering conditions Select the minimum number of system functional indicators that meet the system functional indicator screening criteria. Before selection Each functional indicator is used as a key functional indicator to construct a system functional indicator vector. : ； In the formula, let This is the index of the system function indicator vector. , For the first A vector of system functional indicators.

5. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 4, characterized in that, S2 includes S2.3, a time-delay-dependent degradation model of the control system under the influence of external shocks, based on the effect of shocks on the degradation process: ； ； ； In the formula, and To replace the variable, For the index of impact, , To modify the kernel function, For the first The moment the impact occurred For the first The impact intensity of the second impact, For the impact mapping function, For the shock state coupling function, Let be the Dirac impulse function.

6. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 5, characterized in that, S2 includes S2.4, constructing a reliability model for the underwater vehicle control system: ； In the formula, For reliability, For degradation failure time, For sudden failure time, For the degradation process, The degradation failure threshold, The impact failure threshold, For probability, For any one.

7. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 6, characterized in that, S3 includes, S3.1, constructing non-time-delay parameters. , For impact arrival rate, Impact intensity distribution; Set anomaly detection conditions: ； In the formula, For degenerate incremental sequences, This is the abnormal sensitivity coefficient. Within the sampling time interval The mean, Within the sampling time interval Standard deviation; when If the anomaly detection conditions are met, The sampling interval at which the time point occurs is considered a sudden change interval affected by external shocks. After removing the sudden change intervals, a sample set of samples showing the steady degradation of the underwater control system is obtained. ,exist Simplified system degradation evolution model: ； In the formula, Δ is the sampling time interval. For random disturbance terms, The random disturbance term follows a normal distribution. Let be the variance of the random disturbance term, and be a variable that follows a distribution. The simplified system degradation evolution model is solved using the least squares method to obtain... The estimated value , The estimated value and The estimated value ; Based on the number of impacts and observation time length ,calculate The estimated value : ； based on Determined through statistical distribution fitting And solve using a regression model. If the impact intensity cannot be directly observed, the expectation-maximization algorithm is used to identify the statistical regularity of the impact intensity within the latent variable framework; finally, the initial estimates of the non-time-delay parameters are obtained by integrating the results. ; S3 includes S3.2, constructing time-delay related parameters. , For the first First-order discrete lag coefficient, , For the memory modulation function related to the impact, Index for time delay order, , The time-delay kernel function describes the strength of degenerative memory over continuous time. The time-delay effect intensity coefficient, This is the memory decay coefficient; based on Construct residual sequences : ； In the formula, The impact mapping function obtained during the parameter estimation stage The estimated value, In the current sampling interval The immediate cumulative damage caused by all external shock events occurring within the system to the amount of system degradation; Perform time-delay structure identification on the residual sequence Calculate the sample autocorrelation function : ； In the formula, Total failure time, The sample mean of the residual sequence; Set the threshold for low-order hysteresis terms , For residual sequence With lag order of The degree of direct correlation between time and time, satisfying , The calculation formula is: ； In the formula, The correlation coefficient; Set the time delay order threshold and threshold of high-dimensional redundant lag terms ,when and At that time, the Akaike Information Criterion was introduced to automatically filter out lagging structures: ； In the formula, For the Akaike Information Criteria, The number of time-delay related parameters. For the corresponding lag order Down The likelihood function; By comparing different Below Select to make The smallest lag order is taken as the final lag order; like and It can be linearized and solved using weighted least squares: ； In the formula, Find the parameters that minimize the objective function. This represents the historical degradation state of the discrete-time model; The initial estimates of the time-delay correlation parameters were finally obtained. .

8. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 7, characterized in that, S3 includes S3.3, which introduces an alternating iterative optimization mechanism to... , Initialize as the initial value, and then execute steps S3.3.1, S3.3.2, and S3.3.3 in a loop; S3.3.1, Update the time delay parameters and fix them. Re-estimation using residual sequences ; S3.3.2, Perform non-time-delay parameter correction, to Update known parameters ,right Perform a re-estimation; S3.3.3 Perform convergence determination and set the convergence threshold. and If the relative error of parameter changes between two iterations satisfies: ； ； The estimation result is determined to be convergent, and the following is obtained: and .

9. The method for assessing the reliability of a race-failure control system for an underwater vehicle according to claim 8, characterized in that, S3 includes, S3.4, and will and Substituting the data into the reliability model of the underwater vehicle control system, we obtain the reliability of the underwater vehicle control system.

Images