Abstract
Based on the lack of the variance-based importance measure and the moment-independent one, probability and composite uncertainty importance measures were proposed by imitating the failure probability in the reliability analysis. These indicators looked at the input influence on the shift trend and shift degree of the output distribution shape with the model output skewness and kurtosis. Besides, it was used the state dependent parameter model replacing traditional methods to estimate the moments in these uncertainty importance measures. It reduced the computational cost effectively. Finally, the results of some examples showed the validity and feasibility of the present measure compared with the variance-based importance measure and the moment-independent one. The examples also showed the calculating efficiency and precision of the state dependent parameter model method in comparison with Monte-Carlo simulation.
| Original language | English |
|---|---|
| Pages (from-to) | 64-70+75 |
| Journal | Jisuan Lixue Xuebao/Chinese Journal of Computational Mechanics |
| Volume | 30 |
| Issue number | 1 |
| DOIs | |
| State | Published - Feb 2013 |
Keywords
- Composite importance measure
- Kurtosis
- Probability importance measure
- Sensitivity analysis
- Skewness
- State-dependent parameter estimation