Reliability And Reliability Growth Methods

Based on the available methods, this paper devoted to more extensive researches of reliability methods and reliability growth methods. The main innovations are listed as follows.In order to obtain the failure probability of the implicit limit state equation accurately, the Advanced Mean Value Second Order (AMVSO) method was presented on the basis of the Advanced Mean Value First Order (AMVFO) method. The Advanced Mean Value (AMV) in conjunction with the Response Surface Method (RSM) was also presented to obtain higher precise failure probability of the implicit limit state equation. The results of the AMV in conjunction with the RSM are not sensitive to the positions of the sampling points for determining the response surface equation, which illustrates the robustness of the presented method.A new equivalent method is presented for probabilistic safety analysis of the complex stochastic structure. The presented method provides an incorporation of probabilistic analysis and the standard finite element software. Combined with the powerful mechanics tool, which is familiar to the engineering designer, the presented method can be definitely accepted as an attractive tool for the reliability analysis of the real complex structure.An interval analysis method is presented for confidence interval analysis of random structure response variable, which proposes complicated or implicit response function of basic random variables in general cases. The method could calculate the confidence interval rapidly and accurately when the basic random variables are arbitrary distribution.Due to the more robustness of the confidence interval than that of the point estimation, an interval analysis method for the reliability growth model parameters of the multi-system synchronous development is presented on the basis of the linear regression theory.In order to simplify the reliability growth analysis in the small amount of sample information, Bayes analysis of reliability growth is presented on the basis of Gibbs sampling. As the amount of the sample information is limited, the presented method can give more accurate and simpler evaluation results comparing to the conventional maximum likelihood estimators.