Research On Very High Fatigue Life Distribution And High Reliability Of Structural Components

Since Paris rule was proposed in 1963, there have been no new concepts appearing in the metal fatigue field. But very high cycle fatigue (VHCF) is getting more and more attentions in recent years, key components in modern aircrafts, automobiles and power generation machines often undergo VHCF loading, the cyclic numbers are usually as high as 109~1012. Up to now, little research work has been done for VHCF, and many problems have not been solved completely. Thus the systematical studies on the VHCF life distribution form and the high reliability evaluation method have not only the important significance, but also a practical value.Conditional fatigue limit test for ultralong life and VHCF life distribution test were conducted on smooth funnel shape testpieces of LC4CS aluminum alloy, experiment data and SEM test results show that the VHCF fracture is caused by duplex failure mechanisms. A mixed distribution model is proposed for VHCF and based on it the extreme-based life distribution model is presented for censored life tests which has a more precise distribution in the upper tail life region. Examples are used to demonstrate the new life distribution model and the the censored data analysis method, numerical results show that they have better agreement with the experimental data than the conventional methods.High reliability research is consist of the research of reliability analysis model and the method to describe random variables: According to the definition of safety margin, Reliability analysis models of structural components are proposed based on fatigue residual life and stress intensity factor. The existing attitude that the criterion for choice of empirical frequency function (EFF) in the parameters fitting of variable distributions and in the goodness-of-fit (GOF) tests is arbitrary is rebuked. This dissertation analyzes the skewness and kurtosis characteristics of the failure probability distribution of order statistics, and then proposes a worthwhile EFF based on Beta-distribution. For practical application purposes, a simple approximate formula is also introduced. Additionally, new modified-EFD test statistics and tail test (TT) method are constructed based on the extreme discrepancy criterion. Finally the Monte Carlo method is employed to conduct test power comparisons with a new distribution-free power comparison method. The simulation results show that the TT has a higher power to exclude the wrong distribution hypothesis, which can really decrease the statistical error in the high reliability analysis.