A paper discusses the excessive conservatism that has long been suspected to exist, because of the use of conventional (deterministic) safety factors, in the analysis and design of quasi-static structures. The origin of this conservatism is identified as a violation of statistical error-propagation laws that occurs when statistical data on loads and stresses are reduced to deterministic values and then combined through several computational processes. These findings are suggested to indicate a need to replace deterministic methods with probabilistic methods to prevent violations of error-propagation laws. It is also suggested that, as an alternative to adoption of fully probabilistic methods, it may be more expedient to partially convert deterministic methods to probabilistic ones to retain familiarity, confidence, and correlation with experience.
This work was done by V. Verderaime of Marshall Space Flight Center. MFS-27333