Multi-source data fusion aerospace electronic system reliability evaluation method based on design and manufacturing process constraints

Abstract

The invention discloses a multi-source data fusion aerospace electronic system reliability evaluation method based on design and manufacturing process constraints, and belongs to the technical field of electronic system reliability evaluation. The problem that the reliability evaluation accuracy of the spaceflight electronic system is poor under the condition of incomplete information is solved. The method comprises the following steps: firstly, establishing a digital prototype model of the aerospace electronic system, and establishing a parameter degradation model of key components of the electronic system in combination with a failure physical model so as to obtain a performance degradation reliability function of each key electrical performance parameter of the electronic system; establishing an electronic system function failure propagation network model based on a logical relationship between an electronic system function structure and basic function network nodes, performing state updating on each function network node of the electronic system by adopting a Bayesian inference method, and obtaining a function failure reliability function of each function network node of the electronic system; and then based on the two reliability functions, constructing a system-level reliability model to obtain an overall reliability curve of the electronic system.

Description

Technical Field

[0001] This invention belongs to the field of electronic system reliability assessment technology, specifically relating to a method for assessing the reliability of aerospace electronic systems based on multi-source data fusion. Background Technology

[0002] As spacecraft missions become increasingly complex, their electronic systems are becoming more structurally intricate and functionally coupled, placing higher demands on the reliability assessment of these systems. Existing reliability modeling methods often rely on statistical empirical models or single-source data, making it difficult to comprehensively reflect the operational characteristics and degradation processes of electronic systems. This results in insufficient modeling accuracy and predictive reliability. Particularly in aerospace electronic systems, the sheer number of modules, complex structures, and diverse failure modes, coupled with limited experimental samples and scarce failure data in practical engineering, make traditional methods inadequate for accurate quantitative assessment. Furthermore, aerospace electronic systems suffer from low failure rates and long testing cycles; relying solely on measured data is insufficient to cover key failure modes, while simulation models suffer from incomplete modeling and parameter uncertainties, failing to effectively characterize the impact of component degradation on electronic system performance. While expert knowledge can provide some prior information, a systematic fusion mechanism for quantitative analysis is currently lacking. Therefore, there is an urgent need for a reliability assessment method for aerospace electronic systems that can integrate multi-source information, including electronic system structural models, measured data, degradation patterns, and expert knowledge, to improve assessment accuracy even under conditions of incomplete information. Summary of the Invention

[0003] This invention aims to address the problem of poor accuracy in reliability assessment of aerospace electronic systems under conditions of incomplete information.

[0004] A reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints includes the following steps:

[0005] Step S1: Establish a digital prototype model of the aerospace electronic system that supports parameter degradation and Monte Carlo simulation;

[0006] Step S2: Combine the failure physics model to establish the parameter degradation model of the key components of the electronic system. Based on the digital prototype model constructed in step S1, use the Monte Carlo simulation method to expand the degradation sample of the electronic system.

[0007] Step S3: Using a neural network to fuse limited experimental data with the degradation samples generated in step S2, construct a fused degradation sample set, and based on the failure threshold of key electrical performance parameters, obtain the performance degradation reliability function of each key electrical performance parameter of the electronic system.

[0008] Step S4: Based on the functional structure of the electronic system and the logical relationship between the basic functional network nodes, construct a dynamic logic gate model and establish a functional failure propagation network model of the electronic system.

[0009] Step S5: Combining the online monitoring data of each basic functional network node of the electronic system, and based on the functional failure propagation network constructed in step S4, the Bayesian inference method is used to update the state of each functional network node of the electronic system, and obtain the functional failure reliability function of each functional network node of the electronic system.

[0010] Step S6: Based on the performance degradation reliability function and functional failure reliability function obtained in steps S3 and S5, construct a system-level reliability model, perform series modeling according to the structural and functional relationships between the key modules of the electronic system, and obtain the overall reliability curve of the electronic system.

[0011] Furthermore, the process described in S1 for establishing a digital prototype model of an aerospace electronic system that supports parameter degradation and Monte Carlo simulation includes:

[0012] Step S11: Based on the circuit schematic of the aerospace electronic system, determine the basic functional modules and their functional boundaries of the electronic system, and determine the key electrical performance parameters that affect the lifespan of the electronic system.

[0013] Step S12: Based on the electronic system circuit schematic diagram in step S11, construct the circuit simulation model of each functional module of the electronic system through the Saber simulation platform, and perform single-factor sensitivity analysis. Based on the sensitivity of the circuit output parameters to various components, determine the key components whose parameter degradation directly affects the output electrical performance of the electronic system.

[0014] Step S13: Based on the parameter forms of various key components in step S12, establish a degradation parameter data interface in the circuit simulation model to form a digital prototype model of the aerospace electronic system that supports parameter degradation and Monte Carlo electrical performance simulation.

[0015] Furthermore, step S2, which describes the process of expanding the electronic system degradation sample using the Monte Carlo simulation method, includes:

[0016] Step S21: Combining the results of thermal and electrical stress accelerated life tests, and based on the failure physics model, the degradation model parameters of various key components are fitted by the least squares method to establish parameter degradation models of various key components of the electronic system.

[0017] Step S22: Based on the key component parameter degradation model in step S21, conduct large-scale Monte Carlo sampling to generate multiple sets of samples reflecting the performance degradation of components under the service conditions of electronic systems.

[0018] Step S23: Input the degradation samples from step S22 into the digital prototype model constructed in step S1, and simulate to obtain the degradation trajectory of key electrical performance parameters of the electronic system over time, thereby obtaining a degradation dataset of the electronic system containing the effects of component degradation. This dataset is the degradation dataset of key electrical performance parameters of the electronic system obtained by simulation.

[0019] Furthermore, the specific process of obtaining the performance degradation reliability function of each key electrical performance parameter of the electronic system in step S3 includes:

[0020] Step S31: Based on the electronic system degradation dataset obtained from the simulation in Step S2, a limited number of experimental data are introduced; by splicing the experimental data and the simulation data, a joint matrix that combines the characteristics of both types of data is constructed:

[0021] Let the matrix of the simulated degradation samples be... Where m is the number of samples and n is the number of time points; let the experimental degradation sample vector be... ; Compare y with a vector of length n consisting entirely of 1s The outer product is used to obtain the extended matrix. Then concatenate S and Y along the column directions to obtain the joint matrix. , The matrix is ​​a combination of the characteristics of experimental and simulation data, where S is the matrix of simulation data and Y is the matrix of experimental data.

[0022] Step S32: Using the joint matrix constructed in step S31 as input, the key electrical performance parameter values ​​are output based on the neural network, i.e., the fused degradation samples, to obtain the fused degradation sample set.

[0023] Step S33: Based on the fused degradation samples obtained in step S32, identify the time point when each curve in the sample first reaches the failure threshold according to the failure threshold of the key electrical performance parameters of the electronic system, and form a performance degradation lifetime sample set. Each sample in the sample set corresponds to the "lifetime value" of the key electrical performance parameters of the electronic system.

[0024] By statistically analyzing the lifetime samples, the lifetime distribution function of each key electrical performance parameter from the perspective of performance degradation is obtained, and the performance degradation reliability function of each key electrical performance parameter of the electronic system is further obtained.

[0025] Furthermore, step S4, which describes the process of constructing a dynamic logic gate model and establishing an electronic system functional failure propagation network model based on the functional structure of the electronic system and the logical relationships between the basic functional network nodes, includes:

[0026] Step S41: Based on the circuit schematic of the electronic system, identify the signal transmission and logic dependency relationship of the electronic system, determine the key functional parameters that affect the life of the electronic system, set the basic functional network nodes and intermediate functional network nodes of the electronic system, and construct the functional transmission structure diagram of the electronic system from the basic functional network nodes to the top functional network nodes.

[0027] Step S42: Based on the electronic system functional structure diagram established in step S41 and the logical relationship between each functional network node, construct the electronic system functional failure propagation network model, set dynamic and static logic gates, form a combination structure between the functional states of each network node, and define the propagation path of electronic system functional failure.

[0028] Step S43: Construct a priori reliability model for the basic functional network nodes of the electronic system, and establish reliability functions for the prior information in the priori reliability model respectively;

[0029] Step S44: Input the reliability function of the basic functional network node obtained in step S43 into the electronic system functional failure propagation network model constructed in step S42. Calculate the reliability functions of the intermediate layer functional network node and the top layer functional network node in sequence according to the logic gate combination rules to form the prior reliability distribution of each functional network node in the electronic system. Output the curve of the prior reliability of each functional network node in the electronic system changing with time, i.e., the functional failure reliability function of each functional network node.

[0030] Furthermore, the specific process of obtaining the functional failure reliability function of each functional network node of the electronic system in step S5 includes:

[0031] Step S51: Based on the electronic system functional failure propagation network model constructed in step S4, extract the prior reliability model of each basic functional network node of the electronic system, initialize the initial state of each node, and combine the online operation monitoring platform of the electronic system to collect the operating status and failure time data of each node of the electronic system in real time.

[0032] Step S52: The state of each functional network node of the electronic system is dynamically updated using the Bayesian inference method. The online monitoring data is set as the likelihood function and is input into the prior reliability model in step S51 in real time. The particle filter algorithm is used to realize the iterative correction of the state distribution of the electronic system.

[0033] Step S53: Construct a time-dynamic confidence function mechanism for the operation of the electronic system; model the confidence level based on the state differences of electronic system nodes in the early and later stages of operation to obtain the electronic system confidence function:

[0034]

[0035] Wherein, δ is the rate of change of weight w, which is selected based on the period of online data processing; This is the total duration of the experiment to date. The time when the test failed; This is the confidence decay function, which gradually decreases with increasing runtime. A confidence recovery function that performs confidence repair based on the actual time of failure occurrence;

[0036] Step S54: Based on the updated electronic system state particle set in step S52 and the confidence function in step S53, obtain the posterior reliability function and failure lifetime distribution of each functional network node of the electronic system; plot the reliability change curve, failure probability density curve and confidence curve of each functional network node of the electronic system, and obtain the functional failure reliability function of each functional network node of the electronic system according to the curves.

[0037] Furthermore, the specific process of iteratively correcting the state distribution of the electronic system using the particle filter algorithm in step S52 includes the following steps:

[0038] An initial set of particles is generated based on the prior reliability distribution, where each particle represents an estimate of the reliability of the electronic system at the current time. Noise from the state transition process is introduced to update the particle states. , It has a mean of 0 and a variance of Normal noise;

[0039] The weights of each particle are updated based on monitoring data and observation noise. , The failure time of each functional network node in the electronic system. It is in a particle state; The standard deviation of the observed noise;

[0040] To avoid particle degradation, a resampling operation is performed on the particles of the electronic system to form a new set of particles, which constitutes the posterior probability estimate of the state of the electronic system.

[0041] Furthermore, the confidence decay function gradually decreases with increasing runtime. as follows:

[0042]

[0043] In the formula, This is the total duration of the experiment to date. The estimated node lifetime is based on prior knowledge. This is the descent rate coefficient.

[0044] Confidence recovery function based on actual fault occurrence time as follows:

[0045]

[0046] In the formula, The time when the test failure occurred. For the control parameters of recovery rate, This represents the minimum confidence level.

[0047] Furthermore, the specific process of obtaining the overall reliability curve of the electronic system by performing series modeling according to the structural and functional relationships between the key modules of the electronic system in step S6 includes:

[0048] Step S61: Extract the performance degradation reliability function of each key electrical performance parameter of the electronic system constructed in step S3 and the functional failure reliability function of each functional network node of the electronic system generated in step S5; for each basic functional module of the electronic system, construct the overall reliability function of the module at any time t based on the reliability functions of the key electrical performance parameters and key functional parameters contained in the module; then, perform unified normalization processing on the reliability functions of each module to obtain the reliability function of each module. ;

[0049] Step S62: Based on the electronic system structure diagram and functional logic relationships, identify the connection methods and coupling sequences between modules, and construct an electronic system-level reliability block diagram model; calculate the overall reliability function of the electronic system according to the reliability modeling rules of series electronic systems. Plot reliability change curves to achieve electronic system-level reliability assessment.

[0050] The present invention has the following advantages:

[0051] This invention establishes a dual-path reliability assessment framework that combines degradation modeling and fault propagation analysis. Under conditions of incomplete information, it comprehensively assesses the reliability of electronic systems from two dimensions: performance degradation and functional failure. This not only improves the accuracy of the assessment but also enhances the versatility and practicality of the method, making it applicable to health management and life prediction scenarios for various types of aerospace electronic systems. Attached Figure Description

[0052] Figure 1 This is a flowchart of a reliability assessment method for aerospace electronic systems based on multi-source data fusion constrained by design and manufacturing processes.

[0053] Figure 2 It is a block diagram of the electronic system structure.

[0054] Figure 3 This is a circuit simulation model of the electronic system in the embodiment.

[0055] Figure 4 The key electrical performance parameter U of the electronic system in the embodiment pwr The degenerate distribution.

[0056] Figure 5 The key electrical performance parameter U of the electronic system in the embodiment out The degenerate distribution.

[0057] Figure 6 This refers to the loss changes of key electrical performance parameters of the electronic system during the neural network fusion training process in the embodiment.

[0058] Figure 7 The key electrical performance parameter U of the electronic system in the embodiment pwr Comparison of actual and predicted values ​​after neural network fusion.

[0059] Figure 8 The key electrical performance parameter U of the electronic system in the embodiment out Comparison of actual and predicted values ​​after neural network fusion.

[0060] Figure 9 This is the performance degradation reliability curve of the electronic system in the embodiment.

[0061] Figure 10 This is the functional failure propagation network model of the electronic system in the embodiment.

[0062] Figure 11 This is a time-varying heatmap of the prior functional reliability of each functional network node of the electronic system in the embodiment.

[0063] Figure 12 This refers to the reliability, failure probability density, confidence curve, and failure lifetime of each functional network node of the electronic system when texperiment=2000 in the embodiment.

[0064] Figure 13 It represents the reliability, failure probability density, confidence curve, and failure lifetime of each functional network node of the electronic system under a test duration of 4500 days.

[0065] Figure 14 It represents the reliability, failure probability density, confidence curve, and failure lifetime of each functional network node of the electronic system under a 5000-day test period.

[0066] Figure 15 It represents the reliability, failure probability density, confidence curve, and failure lifetime of each functional network node of the electronic system under a test duration of 5500 days.

[0067] Figure 16 This is the overall reliability curve of the electronic system in the embodiment. Detailed Implementation

[0068] This invention provides a reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints. It is designed for aerospace electronic systems that are characterized by scarce data, complex structures, and difficult-to-reproduce failures. The method comprehensively considers two types of failure mechanisms: performance degradation and functional failure. It conducts reliability assessments for performance degradation and functional failure separately, thereby improving the accuracy and adaptability of the assessment.

[0069] Specific implementation method one: Combining Figure 1 This implementation method is described below.

[0070] This implementation method is a reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints, including the following steps:

[0071] Performance degradation reliability assessment:

[0072] Step S1: Establish a digital prototype model of the aerospace electronic system that supports parameter degradation and Monte Carlo simulation. The specific steps are as follows:

[0073] Step S11: Based on the circuit schematic of the aerospace electronic system, determine the basic functional modules and their functional boundaries of the electronic system, and determine the key electrical performance parameters affecting the lifespan of the electronic system. In some embodiments, the electrical performance parameters mainly include voltage and current. There are many different voltage parameters, including the output voltage of each functional module and the voltage of some intermediate nodes of these functional modules. For example, the output voltage includes the output voltage of the power module, the output voltage of the power distribution module, etc.; the output current also includes the output current of the power module, the output current of the power distribution module, etc. In the following embodiments, two key electrical performance parameters (output voltage of the power module and output voltage of the power module) are selected as examples for demonstration.

[0074] Step S12: Based on the electronic system circuit schematic diagram in step S11, construct the circuit simulation model of each functional module of the electronic system through the Saber simulation platform, and perform single-factor sensitivity analysis. Based on the sensitivity of the circuit output parameters to various components, determine the key components whose parameter degradation directly affects the output electrical performance of the electronic system.

[0075] Sensitivity analysis is used to quantify the impact of component parameter degradation on the output of an electronic system. Assume the performance parameter of the electronic system is T. j The electronic component parameter is p i Then the sensitivity of the electronic system to changes in the performance of this component can be expressed as: ;when When the value is large, it indicates that the first The first component is related to the first If the output characteristics are relatively sensitive, meaning that the component has a significant impact on the performance of the electronic system, it can be identified as a key component.

[0076] Based on the established electronic system circuit simulation model, sensitivity analysis is performed. Under the condition that the component parameters degrade within their lifespan (e.g., ±20%), the parameters of each component in the electronic system circuit simulation model are perturbed one by one, and their impact on the representative performance index U of the electronic system is calculated. sig The influence of (self-test voltage value). U sig As a representative indicator reflecting the overall performance of an electronic system, its amplitude variation is set with a threshold (e.g., ±1%). If the variation caused by component degradation is lower than this threshold, it indicates that the degradation has a small impact on the output of the electronic system and can be regarded as a non-critical component.

[0077] Based on sensitivity analysis, components that have a significant impact on the output parameters of electronic system circuits are identified as key components, such as some DC-DC converters, load resistors, etc. The parameters of these key components (e.g., the capacitance value of a capacitor and the resistance value of a resistor) are the key component parameters.

[0078] Step S13: Based on the parameter forms of various key components in Step S12, establish a degradation parameter data interface in the circuit simulation model to form a digital prototype model of the aerospace electronic system that supports parameter degradation and Monte Carlo electrical performance simulation. The digital prototype model refers to a circuit simulation model that includes the degradation trends of components (meaning the behavior and physical characteristics of the circuit simulation model are the same as the actual situation).

[0079] Step S2: Establish parameter degradation models for key components of the electronic system based on the failure physics model. Using the digital prototype model constructed in Step S1, expand the electronic system degradation sample using Monte Carlo simulation. The specific steps are as follows:

[0080] Step S21: Combining the results of thermal and electrical stress accelerated life tests, and based on the failure physics model, the degradation model parameters of various key components are fitted by the least squares method to establish parameter degradation models of various key components of the electronic system.

[0081] The degradation model of key components takes into account the results of accelerated life tests on various key components under electrical and thermal stress (degradation data of components). Specifically, accelerated life tests are conducted on each type of component in groups under different temperature conditions to obtain time-varying output parameters of the components at different temperatures. By fitting these time-varying output parameters, the values ​​or distributions of parameters in the failure physics model are obtained. The model output is the parameter values ​​or distributions of the failure physics model of various key components.

[0082] Step S22: Based on the key component parameter degradation model in step S21, conduct large-scale Monte Carlo sampling to generate multiple sets of samples reflecting the performance degradation of components under the service conditions of electronic systems.

[0083] Step S23: Input the degradation samples from step S22 into the digital prototype model constructed in step S1, and simulate to obtain the degradation trajectory of key electrical performance parameters of the electronic system over time, thereby obtaining a degradation dataset of the electronic system containing the effects of component degradation. This dataset is the degradation dataset of key electrical performance parameters of the electronic system obtained by simulation.

[0084] Step S3: Based on the multilayer perceptron neural network, the finite experimental data and the degradation samples generated in step S2 are fused to construct a fused degradation sample set. Failure thresholds for key electrical performance parameters are set to obtain the performance degradation reliability function for each key electrical performance parameter of the electronic system. The specific steps are as follows:

[0085] Step S31: Based on the electronic system degradation dataset obtained from the simulation in Step S2, a limited number of experimental data are introduced; by splicing the experimental data and the simulation data, a joint matrix that combines the characteristics of both types of data is constructed, in the following specific form:

[0086] Let the matrix of the simulated degradation samples be... Where m is the number of samples and n is the number of time points; let the experimental degradation sample vector be... .

[0087] Compare y with a vector of length n consisting entirely of 1s. Performing the outer product yields the extended matrix.

[0088] (1)

[0089] Then concatenate S and Y along the column direction to obtain the joint matrix.

[0090] (2)

[0091] in, The matrix is ​​a combination of the characteristics of experimental and simulation data, where S is the matrix of simulation data and Y is the matrix of experimental data.

[0092] Step S32: Based on the multilayer perceptron neural network, the joint matrix constructed in step S31 is used as input, and the key electrical performance parameter values ​​corresponding to the experimental data are used as labels. Nonlinear fitting is used to fuse and train the experimental and simulation data. During training, mean squared error is used as the loss function, the Adam optimizer is employed, and the training and validation sets are divided proportionally.

[0093] Here, "fusion" refers to using a multilayer perceptron (MLP) to learn the nonlinear mapping relationship between a limited number of experimental data and a large number of simulation data, enabling both types of data to be uniformly represented in the feature space, thereby forming a sample of degradation curves for key electrical performance parameters that combine real physical characteristics and simulation trends. The MLP is trained through backpropagation, allowing the network output to simultaneously fit the degradation characteristics of both types of data, thus achieving information complementarity.

[0094] The network input is a joint matrix consisting of an expanded matrix of experimental data y and simulation data S. The expanded matrix, obtained through outer product operations, contains correlation information between the experimental and simulation data in both the time and performance parameter dimensions. The network output is the predicted values ​​of the fused degradation samples, i.e., the degradation curve data points of the key electrical performance parameters of the electronic system under the same time series. These curves include both the full-lifetime trend of the simulation data and the actual degradation characteristics of the measured data, for use in subsequent lifetime statistics.

[0095] During training, the labels are derived from key electrical performance parameter values ​​in the experimental data. Simulation data is primarily used as one of the input features to provide background information on degradation, helping the model integrate features from both experimental and simulation data. Therefore, the electrical performance parameter values ​​from the experimental data are used as labels during training.

[0096] Generation of fused degradation samples: After training, the input for actual fusion is a joint matrix composed of new experimental and simulation data. The prediction results generated by the neural network are the key electrical performance parameter values ​​corresponding to the simulation data that have been fused with the experimental data, i.e., fused degradation samples. These fused degradation samples are obtained based on the feature fusion of experimental and simulation data. Based on the model prediction results, degradation curves are further plotted or fitted.

[0097] Step S33: Based on the fused degradation samples obtained in step S32, set the failure threshold for the key electrical performance parameters of the electronic system, identify the time point when each curve in the samples first reaches the failure threshold, and form a performance degradation lifetime sample set. Each sample in the sample set corresponds to the "lifetime value" (i.e., the time point when the failure threshold is reached) of the key electrical performance parameter of the electronic system. The sample set is a statistical set composed of a large number of such lifetime values. The electronic system has multiple key electrical performance parameters, and each key electrical performance parameter has a performance degradation lifetime sample set here.

[0098] By statistically analyzing the lifetime samples, the lifetime distribution function of each key electrical performance parameter from the perspective of performance degradation is obtained, and the performance degradation reliability function of each key electrical performance parameter of the electronic system is further obtained.

[0099] Functional failure reliability assessment:

[0100] Step S4: Based on the functional structure of the electronic system and the logical relationships between the basic functional network nodes, construct a dynamic logic gate model and establish a functional failure propagation network model for the electronic system. The specific steps are as follows:

[0101] Step S41: Based on the circuit schematic of the electronic system, identify the signal transmission and logic dependency relationship of the electronic system, determine the key functional parameters (digital signals or functional signal parameters, with only two states: normal and failure) that affect the life of the electronic system, set the basic functional network nodes and intermediate functional network nodes of the electronic system, and construct the functional transmission structure diagram of the electronic system from the basic functional network nodes to the top functional network nodes (which ultimately output the functional failure reliability of the electronic system).

[0102] Step S42: Based on the electronic system functional structure diagram established in step S41 and the logical relationships between each functional network node, construct an electronic system functional failure propagation network model, set dynamic and static logic gates such as AND gate, OR gate, PAND gate, SEQ gate, SPARE gate, and FDEP gate to form a combination structure between the functional states of each network node, and define the propagation path of electronic system functional failure;

[0103] Step S43: Construct a priori reliability models for the basic functional network nodes of the electronic system. The prior information in the priori reliability models comes from historical data and expert experience data, including lifetime distribution parameters, mean time between failures (MTBF), failure rate functions, and failure-free test durations, etc. For the above different types of prior data, reliability functions are established using the following modeling methods:

[0104] For data with a known lifetime distribution, the Weibull distribution function is used for modeling.

[0105] For failure-free test duration data, a time-decreasing confidence function based on the arctan function is introduced to form a node reliability change model;

[0106] For other types of data in the prior data, a general form of probability density function is introduced and the reliability is solved by numerical integration;

[0107] Step S44: Input the reliability function of the basic functional network node obtained in step S43 into the electronic system functional failure propagation network model constructed in step S42. Calculate the reliability functions of the intermediate layer functional network node and the top layer functional network node in sequence according to the logic gate combination rules to form the prior reliability distribution of each functional network node in the electronic system. Output the curve of the prior reliability of each functional network node in the electronic system changing with time, i.e., the functional failure reliability function of each functional network node.

[0108] Step S5: Combining the online monitoring data of each basic functional network node of the electronic system, and based on the functional failure propagation network constructed in Step S4, the Bayesian inference method is used to update the state of each functional network node of the electronic system, and obtain the functional failure reliability function of each functional network node of the electronic system. The specific steps are as follows:

[0109] Step S51: Based on the electronic system functional failure propagation network model constructed in step S4, extract the prior reliability model of each basic functional network node of the electronic system, initialize the initial state of each node, and combine the online operation monitoring platform of the electronic system to collect the operating status and failure time data of each node of the electronic system in real time.

[0110] Step S52: The state of each functional network node in the electronic system is dynamically updated using Bayesian inference. The online monitoring data is set as a likelihood function and input in real-time into the prior reliability model in step S51. The particle filter algorithm is then used to iteratively correct the state distribution of the electronic system. This includes the following processing steps:

[0111] An initial set of particles is generated based on the prior reliability distribution, where each particle represents the reliability estimate of the electronic system at the current time. Noise from the state transition process is introduced to update the particle states, satisfying the following formula:

[0112] (3)

[0113] In the formula, It has a mean of 0 and a variance of Normal noise.

[0114] Based on the monitoring data and observation noise, the weight of each particle is updated, and the weight update satisfies the following formula:

[0115] (4)

[0116] in, The failure time of each functional network node in the electronic system. It is obtained from formula (3); This represents the standard deviation of the observed noise.

[0117] To avoid particle degradation, a resampling operation is performed on the particles of the electronic system to form a new set of particles, which constitutes the posterior probability estimate of the state of the electronic system.

[0118] Step S53: Construct a time-dynamic confidence function mechanism for the operation of the electronic system. Two types of functions are introduced to model the confidence level, taking into account the state differences between the electronic system nodes in the early and later stages of operation.

[0119] In the initial stage of operation of the electronic system, when no faults have occurred, the initial reliability of the nodes is set to 1. A confidence decay function is introduced, which gradually decreases as the operating time increases. Its function form is as follows:

[0120] (5)

[0121] In the formula, This is the total duration of the experiment to date. The estimated node lifetime is based on prior knowledge. This is the descent rate coefficient.

[0122] When a node failure occurs in the mid-to-late stages of electronic system operation, a confidence recovery function is introduced to repair the confidence level based on the actual failure occurrence time. The function form is as follows:

[0123] (6)

[0124] In the formula, The time when the test failure occurred. For the control parameters of recovery rate, This represents the minimum confidence level.

[0125] By performing a weighted smooth fusion of the confidence functions of the two stages, a continuous expression of the electronic system's confidence over time is obtained. The fusion function is as follows:

[0126] (7)

[0127] Wherein, δ is the rate of change of weight w, which is selected based on the period of online data processing;

[0128] Step S54: Based on the updated electronic system state particle set in step S52 and the confidence function in step S53, obtain the posterior reliability function and failure lifetime distribution of each functional network node of the electronic system; plot the reliability change curve, failure probability density curve and confidence curve of each functional network node of the electronic system, and obtain the functional failure reliability function of each functional network node of the electronic system according to the curves.

[0129] Step S6: Based on the performance degradation reliability function and functional failure reliability function obtained in steps S3 and S5, construct a system-level reliability model. Perform cascade modeling according to the structural and functional relationships between the key modules of the electronic system to obtain the overall reliability curve of the electronic system. The specific steps are as follows:

[0130] Step S61: Extract the performance degradation reliability functions of each key electrical performance parameter of the electronic system constructed in Step S3 and the functional failure reliability functions of each functional network node of the electronic system generated in Step S5. For each basic functional module of the electronic system, construct the overall reliability function of the module at any time t based on the reliability functions of the key electrical performance parameters and key functional parameters contained in the module. Then, perform unified normalization on the reliability functions of each module to obtain the reliability function of each module. .

[0131] Step S62: Based on the electronic system structure diagram and functional logic relationships, identify the connection methods and coupling sequences between modules, and construct an electronic system-level reliability block diagram model. Calculate the overall reliability function of the electronic system according to the reliability modeling rules for series electronic systems. The formula is as follows:

[0132] (8)

[0133] Plot reliability change curves to achieve electronic system-level reliability assessment.

[0134] Example:

[0135] The engineering object in this embodiment is a certain type of aerospace electronic system. Its main task is to realize on-orbit attitude control and power distribution of spacecraft, and it has the functions of high-precision voltage regulation and multi-channel signal output. The electronic system has obvious modular integration, mainly composed of power supply module, power module, drive module, interface module, main control module, etc., and the modules are tightly coupled through signal links and power supply links. The system takes a DC / DC converter as the core, which undertakes the tasks of voltage regulation and power supply of multiple channels, and realizes closed-loop control through digital controller. The structural block diagram is as follows. Figure 2 As shown.

[0136] By constructing a digital prototype model with degradation modeling capabilities, degradation samples are generated by combining physical degradation simulation and accelerated testing data, and multi-source fusion of performance degradation data is achieved using neural networks. Based on this, a function propagation network model is established, and a prior reliability model of functional nodes is constructed by combining expert experience and historical data. Bayesian inference methods are introduced to achieve online state updates, ultimately obtaining the posterior distribution of the system lifetime, thus enabling a high-reliability assessment of electronic systems under conditions of incomplete information. The process mainly includes the following aspects:

[0137] First, a digital prototype model of the electronic system is established for system performance degradation modeling. Based on the electronic system block diagram and circuit schematic, circuit simulation models of each functional module of the electronic system are constructed using the Saber simulation platform, such as... Figure 3As shown. Based on the system circuit simulation model, the key electrical performance parameters affecting the system lifespan are determined. The output voltage U of the power module is selected. pwr and the power module output voltage U out As representative of the system's key electrical performance parameters, this embodiment uses these two parameters to conduct a reliability assessment and analysis of system performance degradation. Single-factor sensitivity analysis is performed to identify key components whose parameter degradation directly affects system output performance based on the sensitivity of the circuit output parameters to each component. Key system components such as resistors, capacitors, and DC / DC converters are introduced into the failure model using a standard component library and a behavioral modeling language (such as MAST). Simultaneously, to achieve simulation capabilities for component parameter degradation over time, an Arrhenius model and a degradation function interface based on thermal and electrical stress coupling are embedded in the component model, binding the component's operating environment (voltage, current, temperature) to the degradation process, thus forming a dynamically evolving component performance model. Then, the control logic of control sub-modules such as PID controllers, error amplifiers, and PWM modulators is built using the behavioral modeling module in Saber to realize the voltage regulation and closed-loop control functions of the system model. After completing the entire system circuit structure, the system model and the digital prototype are coupled to form an electronic system digital prototype model that supports parameter degradation and Monte Carlo electrical performance simulation.

[0138] Secondly, based on the failure physics model and experimental data, the system performance degradation data sample is expanded. Through system structure analysis, three key components are identified: capacitors, resistors, and DC / DC converters. The parameter degradation of these components will directly affect the system output performance. Based on the Arrhenius model, failure physics models of the key components are constructed, considering the superposition of thermal and electrical stresses, to describe the degradation behavior of component parameters over time. The degradation models of capacitor values, resistor values, and DC / DC converter output voltage are shown in equations (9) to (11), respectively:

[0139] (9)

[0140] (10)

[0141] (11)

[0142] Where C0, R0, and P0 are the initial capacitance, resistance, and output voltage of the component, respectively; V, I, and T are the voltage, current, and temperature stresses that the component withstands, respectively; and E... a Let be the activation energy, k be the Boltzmann constant, and A, B, p, m, and n be undetermined coefficients obtained from the fitting. Accelerated life tests were conducted on key components of the system, and the obtained data were fitted to yield the relevant parameters involved in the model, listed in Table 1.

[0143] Table 1

[0144]

[0145] Based on this, the system degradation sample was expanded. The above model parameters were substituted into the degradation model equations (9) to (11), and 2000 Monte Carlo samplings were performed on each key component to simulate the performance degradation process of the key component parameters during the life cycle. The degradation data was input into the digital prototype model to carry out system-level simulation and obtain U pwr And U out The degradation trends during the component degradation process are as follows: Figure 4 , Figure 5 As shown. By comparing the degradation data obtained from simulation with the theoretical degradation data, it was verified that the root mean square error (RMSE) of the key electrical performance parameters was less than 3.2% and the maximum relative error did not exceed 5.6%, proving the accuracy and degradation mapping capability of the established digital prototype model.

[0146] Then, a neural network is used to fuse simulation and experimental data to complete the reliability assessment of electronic system performance degradation. Based on the simulation degradation dataset, a limited number of experimental data are introduced to construct a multilayer perceptron (MLP) neural network data fusion model. By splicing experimental and simulation data, a joint matrix is ​​constructed, the specific form of which is shown below:

[0147] (12)

[0148] Where y represents experimental data, S represents simulation data, and ⨂ represents the outer product operation, resulting in a matrix. It combines the characteristics of both types of data, and enhances the learning ability of neural networks by introducing additional high-confidence information.

[0149] Using the joint matrix as model input, nonlinear fitting is employed to enhance the model's ability to express experimental characteristics, achieving deep fusion of experimental and simulation data. During training, mean squared error (MSE) is selected as the loss function, the Adam algorithm is used as the optimizer, and the dataset is divided into training and validation sets in a 7:3 ratio. Key system performance parameters U are selected accordingly. pwr And U out Perform fusion modeling training, the training process is as follows: Figure 6 As shown in the figure, the training and validation losses continuously decrease during the iteration process and tend to converge after about the 30th iteration, indicating that the model has good fitting ability and stability.

[0150] Figure 7 and Figure 8 The system's key performance parameters U were displayed. pwr And U outBy comparing the original simulation data with the output of the fused model, it can be seen that the overall trend of the degradation curve after fusion is highly consistent with the experimental observations. The fusion results maintain high accuracy, with quantitative indicators such as RMSE of 0.035V and R² of 0.962, which are better than the predicted performance of the original simulation data (RMSE=0.061V, R²=0.892). The results show that the fused model can effectively eliminate the bias in the simulation data and better approximate the experimental data in terms of the amplitude changes and trends of the simulation data. Based on the degradation curves of the key electrical performance parameters after fusion and their corresponding failure thresholds, the time nodes when the output voltage and current of each channel of the electronic system reach the failure threshold due to performance degradation can be obtained. For the electronic system, once any key electrical performance parameter reaches the failure threshold, it is considered that the system has failed. Therefore, the minimum value of the lifetime of each key electrical performance parameter is selected as the performance degradation lifetime of the electronic system. Statistical fitting of the performance degradation lifetime sample of the fused system shows that its lifetime distribution follows a normal distribution N(5920.84, 174.67). 2 Based on this, the probability density function f(t) of the performance degradation lifetime can be calculated, and the reliability function R(t) of the system based on performance degradation can be further obtained, such as... Figure 9 As shown, when the reliability R(t) = 0.9, the system's performance degradation lifetime is 5697 days.

[0151] Then, based on the logical relationships between the system's functional structure and functional network nodes, a functional failure propagation network model for the electronic system is constructed to achieve prior modeling and quantitative analysis of functional reliability. First, a system-level functional failure propagation network model is constructed. For example... Figure 10 As shown, the network is based on a combination of dynamic and static logic gates. It connects the basic functional network nodes (h, i, j, k, l, m, n) of the system with the top-level functional network node X through intermediate functional network nodes A1, A2, A3, B1, B2, etc., forming a failure path mapping between functional parameters and system output.

[0152] The reliability functions of each basic functional network node are constructed from expert prior data, including lifetime distribution, mean time between failures (MTBF), failure rate function, and failure-free test duration. For different data formats, the following three modeling methods are used to construct the prior reliability functions:

[0153] Given the lifetime distribution data, a two-parameter Weibull distribution function is used for modeling, as shown in equation (13):

[0154] (13)

[0155] In the formula, t is the running time. scale parameter The shape parameter is used to describe the evolution of the system failure rate over time.

[0156] For failure-free test duration data: For small samples or newly developed components, the initial reliability is assumed to be 1, and an arctan function that decreases with time is introduced as the node confidence factor to simulate the weakening of prior reliability confidence as the running time increases.

[0157] For other types of data in the prior data: a general failure probability density function is introduced, in the form of equation (14):

[0158] (14)

[0159] The cumulative failure rate distribution function is obtained through numerical integration, and then the reliability function is calculated.

[0160] The aforementioned prior modeling provides the basic parameter inputs for the system functional failure propagation network model. In the functional failure propagation network model, each basic functional network node is combined into an intermediate layer functional network node through logic gates, and its reliability is determined by the type of logic gate. The reliability calculation formulas for various types of logic gates are listed in Table 2 below:

[0161] Table 2

[0162]

[0163] in, For the reliability of basic network nodes; the PAND gate assumes two basic events A and B, requiring A to fail before B; the SEQ gate requires that n events fail sequentially. The indicator function; the SPARE gate describes the spare structure, where, The reliability of the main components, The failure rate of the main component at time s. After the spare parts are connected, Time reliability; FDEP gates describe situations where a triggering event causes a series of dependent events to fail. To ensure the reliability of the event triggering, Let be the probability that the i-th dependent event will inevitably fail after being triggered. Let be the independent reliability of the i-th dependent event.

[0164] The above mathematical expressions form the basis for reliability propagation from the component level to the system output level. The reliability functions of each basic functional network node are input into the system functional failure propagation network model. Following the logic gate combination rules, the reliability functions of intermediate-level and top-level functional network nodes are calculated sequentially, forming the prior reliability distribution curves of each functional network node in the electronic system. Figure 11 The heatmaps show the prior reliability of each functional network node of the electronic system at different times, reflecting the temporal evolution trend of the system's functional reliability.

[0165] Then, Bayesian inference is used to update the state of each functional network node, obtaining the posterior distribution of functional failure lifetime and reliability curves for each node in the electronic system. Based on the system's prior reliability model, online measured data gradually plays a dominant role as the system's operating time increases. By collecting system operating status information in real time, online observations of each basic functional network node and intermediate layer functional network node can be obtained. Using Bayesian inference, the online monitoring data is used as a likelihood function to correct the prior distribution, resulting in a posterior distribution that closely reflects actual operating conditions.

[0166] The Bayesian update mechanism achieves dynamic correction of the system state through particle filtering. The core idea of ​​particle filtering is to use a large number of random particles to represent the possible states of the system at different time points. Its overall process includes the following stages:

[0167] Parameter initialization: Generate an initial particle set based on the prior distribution, where each particle represents the state of the system at time t.

[0168] As time progresses, the particle state is updated due to process noise, as shown in equation (15):

[0169] (15)

[0170] The noise values ​​follow a pattern with a mean of 0 and a variance of . It follows a normal distribution.

[0171] Weight Update: When a basic functional network node fails during real-time monitoring, the failure time is recorded, and the node failure time is calculated based on the conversion ratio between hot standby and actual operating conditions in the accelerated test. The particle weights are updated according to the observed value y, as shown in Equation (16):

[0172] (16)

[0173] in The standard deviation of the observed noise.

[0174] Resampling mechanism: To avoid weight degradation, particle resampling is performed to make high-weight particles dominate.

[0175] This dynamic Bayesian process supports continuous prediction in fault-free conditions and rapid after-fault updates, improving the model's accuracy in representing the system state. To more realistically reflect the model's confidence level at different stages, a dynamically adjusted confidence function mechanism is designed:

[0176] Initial stage: The system is still in a fault-free state. To avoid over-updating due to occasional fluctuations, the prior reliability function is kept unchanged, but the confidence value decreases over time. The change law is described by the arctan function, as shown in equation (17):

[0177] (17)

[0178] In the formula, This is the current total test duration. To estimate node lifetime, This is the descent rate coefficient.

[0179] In the mid-to-late stage: nodes begin to fail, and the confidence level is improved by using an exponential recovery function, as shown in equation (18):

[0180] (18)

[0181] In the formula, The time of the node's first failure. For recovery rate.

[0182] Confidence score fusion: For the confidence scores from different time periods, the Logistic function is used to smoothly fuse the confidence scores, as shown in equation (19):

[0183] (19)

[0184] The δ fusion speed control parameter is set according to the online data processing cycle.

[0185] The aforementioned function constitutes a dynamic adjustment mechanism for the transition from the prior stage to the posterior stage. It lowers the model confidence level and improves predictive conservatism when actual fault information is lacking; and quickly restores the confidence level and enhances the reliability of model predictions after an actual fault is observed.

[0186] Based on the constructed particle filter and confidence fusion mechanism, time-related prior and posterior information are jointly introduced into the system-level reliability assessment architecture. Let t = It can calculate the reliability function, failure probability density function, and confidence curve of the top-level functional network nodes and each intermediate-level functional network node of the system, such as... Figure 12 As shown in the figure, X represents the top-level functional network node of the system, A1, A3, B2, B3, and B6 represent the intermediate-level functional network nodes, and i, j, and k represent the basic functional network nodes. This figure illustrates the degradation trend of the overall system performance over time, as well as the nonlinear effects of logic gate combinations.

[0187] like Figure 13 , Figure 14 and Figure 15 As shown, under different test durations (e.g., 4500, 5000, and 5500 days), the reliability curves of the basic functional network node (node ​​i), intermediate functional network nodes (nodes A3 and B3), and top-level functional network node (node ​​X) exhibit a trend of first decreasing and then increasing. The confidence level increases with the accumulation of a priori information, but due to the uncertainty of the system structure, the overall confidence level of the top-level functional network node X is relatively low. Taking the system reliability falling to 0.9 as the failure criterion, the model predicts that the system will fail at 3012 days on day 1000. As the test progresses, the model is gradually corrected, and finally, at day 5500, the failure time of all node signals is identified, and the system functional failure lifetime is judged to be 4067 days. At this time, the overall confidence level is close to 1.

[0188] Finally, based on the reliability assessment results of performance degradation and functional failure, the overall system reliability assessment result is obtained. The performance degradation reliability function constructed earlier and the posterior reliability functions of the generated system functional network nodes are extracted. Time reliability expressions for the key functional modules of the electronic system are established separately according to module division, and the reliability functions of each module at any time t are constructed. The data is then standardized and normalized. Based on the electronic system structure diagram and functional logic relationships, the connection methods and coupling sequences between modules are identified, and a system-level reliability block diagram model is constructed. Following the reliability modeling rules for series systems, the overall system reliability function is calculated. The formula is as follows:

[0189] (20)

[0190] Plot the reliability variation curve to obtain the overall system reliability assessment result, such as... Figure 16 It can be seen that when the reliability is 0.9, the overall lifespan of the system is 4067 days.

[0191] The above examples of the present invention are merely illustrative of the computational model and process of the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is impossible to exhaustively list all possible implementations here. Any obvious variations or modifications derived from the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints, characterized in that, Includes the following steps: Step S1: Establish a digital prototype model of the aerospace electronic system that supports parameter degradation and Monte Carlo simulation; Step S2: Combine the failure physics model to establish the parameter degradation model of the key components of the electronic system. Based on the digital prototype model constructed in step S1, use the Monte Carlo simulation method to expand the degradation sample of the electronic system. Step S3: Using a neural network to fuse limited experimental data with the degradation samples generated in step S2, construct a fused degradation sample set, and based on the failure threshold of key electrical performance parameters, obtain the performance degradation reliability function of each key electrical performance parameter of the electronic system. Step S4: Based on the functional structure of the electronic system and the logical relationships between the basic functional network nodes, construct a dynamic logic gate model and establish a functional failure propagation network model for the electronic system. Step S5: Combining the online monitoring data of each basic functional network node of the electronic system, and based on the functional failure propagation network constructed in step S4, the Bayesian inference method is used to update the state of each functional network node of the electronic system, and obtain the functional failure reliability function of each functional network node of the electronic system. Step S6: Based on the performance degradation reliability function and functional failure reliability function obtained in steps S3 and S5, construct a system-level reliability model, perform series modeling according to the structural and functional relationships between the key modules of the electronic system, and obtain the overall reliability curve of the electronic system.

2. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 1, characterized in that, The process of establishing a digital prototype model of an aerospace electronic system that supports parameter degradation and Monte Carlo simulation, as described in S1, includes: Step S11: Based on the circuit schematic of the aerospace electronic system, determine the basic functional modules and their functional boundaries of the electronic system, and determine the key electrical performance parameters that affect the lifespan of the electronic system. Step S12: Based on the electronic system circuit schematic diagram in step S11, construct the circuit simulation model of each functional module of the electronic system through the Saber simulation platform, and perform single-factor sensitivity analysis. Based on the sensitivity of the circuit output parameters to various components, determine the key components whose parameter degradation directly affects the output electrical performance of the electronic system. Step S13: Based on the parameter forms of various key components in step S12, establish a degradation parameter data interface in the circuit simulation model to form a digital prototype model of the aerospace electronic system that supports parameter degradation and Monte Carlo electrical performance simulation.

3. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 1, characterized in that, Step S2, which describes the process of expanding the electronic system degradation sample using the Monte Carlo simulation method, includes: Step S21: Combining the results of thermal and electrical stress accelerated life tests, and based on the failure physics model, the degradation model parameters of various key components are fitted by the least squares method to establish parameter degradation models of various key components of the electronic system. Step S22: Based on the key component parameter degradation model in step S21, conduct large-scale Monte Carlo sampling to generate multiple sets of samples reflecting the performance degradation of components under the service conditions of electronic systems. Step S23: Input the degradation samples from step S22 into the digital prototype model constructed in step S1, and simulate to obtain the degradation trajectory of key electrical performance parameters of the electronic system over time, thereby obtaining a degradation dataset of the electronic system containing the effects of component degradation. This dataset is the degradation dataset of key electrical performance parameters of the electronic system obtained by simulation.

4. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 1, characterized in that, The specific process of obtaining the performance degradation reliability function of each key electrical performance parameter of the electronic system in step S3 includes: Step S31: Based on the electronic system degradation dataset obtained from the simulation in Step S2, a limited number of experimental data are introduced; by splicing the experimental data and the simulation data, a joint matrix that combines the characteristics of both types of data is constructed: Let the matrix of the simulated degradation samples be... Where m is the number of samples and n is the number of time points; let the experimental degradation sample vector be... ; Compare y with a vector of length n consisting entirely of 1s The outer product is used to obtain the extended matrix. Then concatenate S and Y along the column directions to obtain the joint matrix. , The matrix is ​​a combination of the characteristics of experimental and simulation data, where S is the matrix of simulation data and Y is the matrix of experimental data. Step S32: Using the joint matrix constructed in step S31 as input, the key electrical performance parameter values ​​are output based on the neural network, i.e., the fused degradation samples, to obtain the fused degradation sample set. Step S33: Based on the fused degradation samples obtained in step S32, identify the time point when each curve in the sample first reaches the failure threshold according to the failure threshold of the key electrical performance parameters of the electronic system, and form a performance degradation lifetime sample set. Each sample in the sample set corresponds to the "lifetime value" of the key electrical performance parameters of the electronic system. By statistically analyzing the lifetime samples, the lifetime distribution function of each key electrical performance parameter from the perspective of performance degradation is obtained, and the performance degradation reliability function of each key electrical performance parameter of the electronic system is further obtained.

5. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 1, characterized in that, Step S4 describes the process of constructing a dynamic logic gate model and establishing an electronic system functional failure propagation network model based on the functional structure of the electronic system and the logical relationships between the basic functional network nodes. Step S41: Based on the circuit schematic of the electronic system, identify the signal transmission and logic dependency relationship of the electronic system, determine the key functional parameters that affect the life of the electronic system, set the basic functional network nodes and intermediate functional network nodes of the electronic system, and construct the functional transmission structure diagram of the electronic system from the basic functional network nodes to the top functional network nodes. Step S42: Based on the electronic system functional structure diagram established in step S41 and the logical relationship between each functional network node, construct the electronic system functional failure propagation network model, set dynamic and static logic gates, form a combination structure between the functional states of each network node, and define the propagation path of electronic system functional failure. Step S43: Construct a priori reliability model for the basic functional network nodes of the electronic system, and establish reliability functions for the prior information in the priori reliability model respectively; Step S44: Input the reliability function of the basic functional network node obtained in step S43 into the electronic system functional failure propagation network model constructed in step S42. Calculate the reliability functions of the intermediate layer functional network node and the top layer functional network node in sequence according to the logic gate combination rules to form the prior reliability distribution of each functional network node in the electronic system. Output the curve of the prior reliability of each functional network node in the electronic system changing with time, i.e., the functional failure reliability function of each functional network node.

6. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 1, characterized in that, The specific process of obtaining the functional failure reliability function of each functional network node of the electronic system in step S5 includes: Step S51: Based on the electronic system functional failure propagation network model constructed in step S4, extract the prior reliability model of each basic functional network node of the electronic system, initialize the initial state of each node, and combine the online operation monitoring platform of the electronic system to collect the operating status and failure time data of each node of the electronic system in real time. Step S52: The state of each functional network node of the electronic system is dynamically updated using the Bayesian inference method. The online monitoring data is set as the likelihood function and is input into the prior reliability model in step S51 in real time. The particle filter algorithm is used to realize the iterative correction of the state distribution of the electronic system. Step S53: Construct a time-dynamic confidence function mechanism for the operation of the electronic system; model the confidence level based on the state differences of electronic system nodes in the early and later stages of operation to obtain the electronic system confidence function: Wherein, δ is the rate of change of weight w, which is selected based on the period of online data processing; This is the total duration of the experiment to date. The time when the test failed; This is the confidence decay function, which gradually decreases with increasing runtime. A confidence recovery function that performs confidence repair based on the actual time of failure occurrence; Step S54: Based on the updated electronic system state particle set in step S52 and the confidence function in step S53, obtain the posterior reliability function and failure lifetime distribution of each functional network node of the electronic system; plot the reliability change curve, failure probability density curve and confidence curve of each functional network node of the electronic system, and obtain the functional failure reliability function of each functional network node of the electronic system according to the curves.

7. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 6, characterized in that, The specific process of iteratively correcting the state distribution of the electronic system using the particle filter algorithm, as described in step S52, includes the following steps: An initial set of particles is generated based on the prior reliability distribution, where each particle represents the reliability estimate of the electronic system at the current time. Noise from the state transition process is introduced to update the particle states. , It has a mean of 0 and a variance of Normal noise; The weights of each particle are updated based on monitoring data and observation noise. , The failure time of each functional network node in the electronic system. It is in a particle state; The standard deviation of the observed noise; To avoid particle degradation, a resampling operation is performed on the particles of the electronic system to form a new set of particles, which constitutes the posterior probability estimate of the state of the electronic system.

8. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 6, characterized in that, Confidence decay function that gradually decreases with runtime as follows: In the formula, This is the total duration of the experiment to date. The estimated node lifetime is based on prior knowledge. This is the descent rate coefficient.

9. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints as described in claim 6, characterized in that, Confidence recovery function based on actual fault occurrence time as follows: In the formula, The time when the test failure occurred. For the control parameters of recovery rate, This represents the minimum confidence level.

10. The reliability assessment method for aerospace electronic systems based on multi-source data fusion under design and manufacturing process constraints according to claim 1, characterized in that, Step S6, which describes the process of performing a series modeling based on the structural and functional relationships between the key modules of the electronic system to obtain the overall reliability curve of the electronic system, includes: Step S61: Extract the performance degradation reliability function of each key electrical performance parameter of the electronic system constructed in step S3 and the functional failure reliability function of each functional network node of the electronic system generated in step S5; for each basic functional module of the electronic system, construct the overall reliability function of the module at any time t based on the reliability functions of the key electrical performance parameters and key functional parameters contained in the module; then, perform unified normalization processing on the reliability functions of each module to obtain the reliability function of each module. ; Step S62: Based on the electronic system structure diagram and functional logic relationships, identify the connection methods and coupling sequences between modules, and construct an electronic system-level reliability block diagram model; calculate the overall reliability function of the electronic system according to the reliability modeling rules of series electronic systems. Plot reliability change curves to achieve electronic system-level reliability assessment.

Images