Considerations of environmental effects in fatigue have resulted in proposals to augment the original basis and deterministic methods of fatigue assessment, such as in the ASME Code Subarticles NB-3200 and NB-3600. This process of deterministically combining various elements, each with its own set of safety factors, of the CUF-based assessment has the potential to be overly conservative and restrictive in practice. Furthermore, as in the commonly used deterministic design approaches, fixed safety factors are subjectively assigned; as a result the approach does not provide a logical basis to account for uncertainties or variability, and the resulting level of reliability cannot be assessed quantitatively. Therefore, it is useful and desirable to complement the simplicity of deterministic approach by relating the safety factor to target reliability (or probability of meeting the design criterion) so that an appropriately adequate conservatism can be utilized. The feasibility of such an inter-relation and its underlying probabilistic basis were demonstrated in our recent paper that provided a rational basis to account for the significant uncertainties in assessing the CUF-based fatigue including environmental effects.
The objective of this paper is to provide further assessment of the above basis for uncertainty quantification and its linkage to deterministic safety factor approach, with additional focus on the quantitative sensitivity analysis of varied sources of uncertainty in the CUF estimation. Results of the case studies implementing the proposed approach combining these uncertainties are presented. New expressions for sensitivity assessment are developed. Results of sensitivity analysis are presented with the goal of demonstrating the sensitivity/ranking of significant contributors to the final CUF uncertainty and the resulting deterministic margins in relation to the target (specified) reliability. The overall approach, also summarized in the paper, utilizes the generally accepted concept of propagation of input uncertainties based on the Taylor series method and the framework of the stress-strength interference technique. The utility and limitations of the approach are discussed in defining the acceptable deterministic margins and in quantifying the impact of various elements of conservatism in the current CUF based fatigue evaluations.