Reliability analysis and risk assessment for structural system design and certification.

摘要

The certification of mechanical systems for deployment requires testing and demonstration of a specified level of reliability with a specified level of confidence. A new methodology is proposed for system certification. This method combines the analytical reliability methods with testing to provide a more efficient, cost-effective and streamlined design process. A new system reliability analysis algorithm is developed to account for the effects of multiple interactive failure modes. Several new techniques to compute the confidence in the predicted reliability using the uncertainty in distribution parameters of various design variables are presented. A method based on Bayesian statistics is developed which uses the experimental data gathered at various levels of system development to calibrate the confidence intervals on the predicted reliability for the effects of various approximations and errors made during the design process. Lastly, an algorithm is developed to compute the effect of this calibration on the distribution parameters of the design variables.