| The management of airlines safe status and maintenance quality is the important content of Civil Aviation Management.It ought to make comprehensive and objective identification and evaluation of the maintenance system.Then adjustment to the sub-systems with hidden dangers or weakness may keep overall safe level within acceptable limits.And the prediction is signif icant to improve the reliability and safety of the system.On account of the shortcomings in the previous study,the research contents were proposed from three aspects:Firstly,to settle the safety assessment and improvement problems,the airlines maintenance quality assessment model was built based on Bayesian Net,Fuzzy Method and interval mathematics.Risk factors were analyzed due to the Bayesian inference.Then,the nodes were reduced based on the computation of mutual information.Combined with random set theory,the combined factors were also analyzed with both forward diagnosis and backward reasoning.Secondly,to deal with less labeled data,large number of unlabeled samples and a small number of labeled samples were effectively and reasonably utilized.Based on semi-supervised Bayes ian neural network,the relationship between the index and the safety status was trained.Making use of the SOBOL method,the sensitivity of each index was analyzed by substituting the actual operation data.Through the sensitivity index and regulation analys is,the influence of different factors and the effect each other were analyzed and the corresponding adjustment measure was proposed.Finally,aiming at the early warning of the maintenance system,the samples were expanded based on semi-supervised method.Then the samples were trained in the Deep belief network and the classification of the risk was predicted.The expected training effect was obtained through the adjustment of the parameters.The experiment results showed that the depth belief network can effectively warn the maintenance quality management system. |
PDF Full Text Request