Search

CN-122020860-A - Method for updating time-varying reliability of aircraft movement mechanism based on failure probability function

CN122020860ACN 122020860 ACN122020860 ACN 122020860ACN-122020860-A

Abstract

The invention discloses an aircraft movement mechanism time-varying reliability updating method based on a failure probability function. Firstly, constructing a layered Bayesian network model according to the functional principle and the fault transfer path of an airplane retraction jack system, and learning network parameters by utilizing training samples generated by a physical simulation model. Secondly, sampling from prior distribution of design variables, and calculating failure probability of the multi-failure mode mechanism system by combining the Bayesian network and the subset simulation-Markov chain Monte Carlo method, so as to establish a function model of the failure probability of the system with respect to design variable distribution parameters. Finally, the distribution parameters of the design variables are updated according to the observed data continuously acquired in the actual operation of the mechanism system, and the failure probability function is directly utilized to update the failure probability estimated value rapidly. The invention realizes the reliability prediction based on priori knowledge and the dynamic update based on measured data, and can effectively support the reliability analysis and decision of the mechanism system in different stages of the whole life cycle.

Inventors

  • LI WEI
  • LI XINYAO
  • JIN YAN
  • ZHANG HAOYU
  • LI HONGSHUANG

Assignees

  • 南京航空航天大学

Dates

Publication Date
20260512
Application Date
20260410

Claims (10)

  1. 1. The method for updating the time-varying reliability of the aircraft movement mechanism based on the failure probability function is characterized by comprising the following steps of: Step 1, performing failure mode analysis on an aircraft motion mechanism system, and establishing a layered Bayesian causal network according to the functional principle of the system and a fault transmission path, wherein the network at least comprises a design variable layer, a failure mode functional function layer and a system failure layer so as to describe causal transmission relations from the uncertainty of the design variable to a specific failure mode and then to the system failure; Step 2, generating a training sample based on a physical simulation model of the aircraft motion mechanism, and establishing a proxy model for each failure mode in the failure mode functional function layer by using the training sample as a conditional probability distribution of a corresponding node in a Bayesian network so as to obtain a parameterized hierarchical Bayesian network model; step3, taking the parameterized hierarchical Bayesian network model as a calculation core, and aiming at design variable distribution parameters A series of discrete values in the value interval are calculated by adopting a method combining subset simulation and Markov chain Monte Carlo, and the corresponding system failure probability is calculated respectively Based on the obtained multiple groups Data, using interpolation, regression or machine learning methods, constructs parameters distributed from design variables Probability of failure to system Explicit function model of (c) ; Step 4, inputting the continuously collected observation data in the actual running process of the aircraft movement mechanism into the parameterized layer Bayesian network model, and updating the distribution parameters of the design variable layer through backward reasoning to obtain updated parameters And (3) updating the updated distribution parameters Directly substituting the explicit function model And outputting the updated system failure probability in real time.
  2. 2. The method of claim 1, wherein in step 1, the hierarchical bayesian network model further comprises a design variable distribution parameter layer located below the design variable layer.
  3. 3. The method of claim 1, wherein in step 2, the proxy model is built by extracting multiple input samples X in a priori distribution space of design variables by Latin hypercube sampling method to cover all possible value ranges of variables, inputting the samples one by one into a dynamic simulation model for characterizing physical mechanism of a mechanism to simulate and obtain real engineering response, calculating function values g corresponding to the samples according to the output response and criterion formulas of failure modes, and according to data sets for each failure mode A proxy model is built to describe the mapping from X to g.
  4. 4. A method according to claim 3, wherein in step 2, the proxy model is a gaussian process regression model.
  5. 5. The method of claim 1, wherein in step3, parameters are distributed for design variables The series of discrete values in the value interval refers to that the value range of the distribution parameter of each design variable is uniformly or non-uniformly discretized to form a discrete grid point set in the multidimensional parameter space.
  6. 6. The method according to any one of claims 1-5, wherein in step 3, the method of computing the system failure probability using subset simulation in combination with markov chain monte carlo comprises: step 3.1, hierarchical setting, namely, the system failure domain is set Divided into M layers of intermediate failure domain sequences In which, in the process, As a function of the system, When the system is in a failure state, For the kth time step, subscript i=1, 2..m represents an i-th layer intermediate failure domain; step 3.2, calculating first layer probability, namely estimating that a sample falls into a first layer intermediate failure domain through Monte Carlo sampling Probability of (2) ; Step 3.3, serializing the conditional probability estimate i=2 for the i-th layer, M to fall into the previous layer failure domain Is used as seed, and a Markov chain Monte Carlo method is used for the seed Intra-generating a sample chain and estimating conditional probabilities Outputting an estimated value of the conditional probability ; Step 3.4, probability synthesis, namely leading the first layer probability to be Probability of each condition ,..., Multiplying to obtain the failure probability of the system 。
  7. 7. The method according to claim 6, wherein the step 3.2 is performed by Random extraction in design variable probability distribution as parameter Individual samples Each of is provided with Inputting the system function value into the parameterized hierarchical Bayesian network in the step 2 for forward reasoning to obtain the system function value By indicating a function Judging whether the sample falls into I.e. Wherein the function is indicated 1 In the domain, otherwise 0, counting the number of samples falling into the domain to obtain Estimate of (2) 。
  8. 8. The method according to claim 6, wherein in step 3.3, the markov chain monte carlo method is: step 3.31 from the previous layer Selecting one of the samples of (a) as the starting point of the Markov chain from the current sample point on the current chain Starting, generating a candidate state according to a preset suggestion distribution ; Step 3.32, calculating the scaling factor of the candidate receiving state Wherein Probability density functions for design variables; Step 3.33, accepting the candidate state with probability min (1, r) according to the accept-reject policy As the next state of the chain Otherwise keep state Unchanged, repeating steps 3.31 to 3.33 to generate a sufficient number of Markov chain states to form a distribution from the condition In the sample set, statistics fall into the region Outputs an estimate of conditional probability , 。
  9. 9. The method according to claim 1, wherein in step 4, the updating of the distribution parameters of the design variable layer by backward reasoning means that the posterior probability distribution of the design variable distribution parameters is calculated using bayesian theorem, taking the observed data as a condition, and taking the mean or mode of the posterior distribution as the updated parameters 。
  10. 10. The method of claim 1, wherein the failure mode comprises a mechanism movement stuck failure and/or a mechanism movement accuracy failure.

Description

Method for updating time-varying reliability of aircraft movement mechanism based on failure probability function Technical Field The invention belongs to the field of reliability analysis of a motion mechanism, and particularly relates to a time-varying reliability updating method of an aircraft motion mechanism based on a Bayesian network and a failure probability function. Background The aircraft motion mechanism system, such as a retraction system of a flap, a slat, a landing gear and a speed reducing plate, is a key complex mechanical subsystem for guaranteeing the flight performance and safety of an aircraft. These mechanisms require precise controlled movements at each stage of flight, the reliability of which directly affects flight safety and aircraft integrity. However, due to the high integration and complex failure mechanisms, such as wear, fatigue, motion jamming, and motion precision failure, the failure modes coexist, and the coupling effect of variable loads and environments is received, the comprehensive, accurate and efficient evaluation of the reliability of the device faces a series of serious challenges. At present, the reliability analysis method for the mechanism in engineering mainly has the following limitations that firstly, the traditional methods such as fault tree, FMEA and the like are difficult to accurately describe complex coupling relations and fault transmission paths among multiple failure modes, and modeling capability is insufficient. Secondly, when evaluating safety critical events involving very small failure probabilities, the standard Monte Carlo simulation method requires a large number of samples, and is high in calculation cost and low in efficiency. Furthermore, most existing methods severely rely on limited historical fault data or test data for statistical inference, but failure data books in the whole life cycle of the mechanism are sparse and have high acquisition cost, and a simulation model reflecting a fault physical mechanism is difficult to fuse by a pure data driving method, so that the physical credibility and extrapolation of an evaluation result are limited. More importantly, the existing framework is usually static cured, and cannot be effectively integrated with state monitoring and performance observation data continuously generated in the actual running process of the system, so that dynamic updating and calibration of reliability prediction along with service state degradation cannot be realized, and predictive maintenance and real-time risk decision support are difficult. Disclosure of Invention In order to overcome the technical problems of high calculation cost, low solving efficiency of small failure probability and incapability of fusing real-time data to perform dynamic update in the conventional reliability evaluation method of the aircraft movement mechanism, the invention provides a dynamic reliability analysis method based on a Bayesian network and an integrated failure probability function, and aims to realize efficient and accurate evaluation and online update of the reliability of the whole life cycle of the mechanism. The technical scheme is characterized in that a hierarchical Bayesian network is constructed to fuse a physical mechanism and uncertainty, a high-efficiency probability algorithm is adopted to solve the small failure probability, and finally an explicit function model of the failure probability is established to realize quick dynamic updating. The method specifically comprises a model construction step (i.e. step 1-3) and a dynamic updating step (i.e. step 4), and specifically comprises the following steps: Step 1, performing failure mode analysis on an aircraft motion mechanism system, and establishing a layered Bayesian causal network according to the functional principle of the system and a fault transmission path, wherein the network at least comprises a design variable layer, a failure mode functional function layer and a system failure layer so as to describe causal transmission relations from the uncertainty of the design variable to a specific failure mode and then to the system failure; Step 2, generating a training sample based on a physical simulation model of the aircraft motion mechanism, and establishing a proxy model for each failure mode in the failure mode function layer by using the training sample to serve as conditional probability distribution of corresponding nodes in the Bayesian network so as to obtain a parameterized hierarchical Bayesian network model; step3, taking the parameterized hierarchical Bayesian network model as a calculation core, and aiming at design variable distribution parameters A series of discrete values in the value interval are calculated by adopting a method combining subset simulation and Markov chain Monte Carlo, and the corresponding system failure probability is calculated respectivelyBased on the obtained multiple groupsData, using interpolation, regression or machi