Search for Author, Title, Keyword
Global non-probabilistic reliability sensitivity analysis based on surrogate model
More details
Hide details
Online publication date: 2022-09-05
Publication date: 2022-09-05
Eksploatacja i Niezawodność – Maintenance and Reliability 2022;24(4):612–616
  • Global non-probabilistic sensitivity analysis method is proposed.
  • The proposed method considers main effect of interval variables.
  • The proposed method is easy to use and does not require probability distribution.
Sensitivity analysis is used to find the key variables which have significant effect on system reliability. For a product in early design stage, it is impossible to collect sufficient samples. Thus, the probabilistic-based reliability sensitivity analysis methods are difficult to use due to the requirement of probability distribution. As an alternative, interval can be used because it only requires few samples. In this study, an effective global non-probabilistic sensitivity analysis based on adaptive Kriging model is proposed. The global accuracy Kriging model is constructed to reduce overall computational cost. Subsequently, the global non-probabilistic sensitivity analysis method is developed. Compared to existing non-probabilistic sensitivity analysis methods, the proposed method is a global non-probabilistic reliability sensitivity analysis method. The proposed method is easy to use and does not require probability distribution of the input variables. The applicability of proposed method is demonstrated via two examples.