Application of Survival Analysis Techniques to Probabilistic Assessment of Fatigue in Steel Bridges

Abstract

The fatigue of engineering materials under repetitive loading is a significant issue affecting the design and durability of components and systems in a variety of engineering-related applications including civil, mechanical, aerospace, automotive, and electronics. Many factors can affect the service life of a component or system under repetitive loading, such as the type of structure, loading, connection details, stress state, peak stress or stress range, surface condition, temperature, and environmental exposure. Currently, there is no comprehensive probabilistic approach that can systematically address all the factors that contribute to fatigue on a single mathematical platform. However, advanced analysis techniques developed for and used in various medical research applications may hold some answers. In such research, probabilistic assessments of time to reach a milestone (e.g., time to recurrence of a disease) is considered under the influence of a range of numerical and/or categorical parameters. The experimental data obtained from observations during research is used to generate the analysis models. Such “survival analysis” involves comprehensive, multi-parameter nonlinear regression techniques that incorporate various baseline statistical distributions. This research aims to develop, apply, and verify long-standing survival analysis techniques, widely used in medical research, to the engineering fatigue problem. This research will also use conditional survival analysis techniques derived from the conditional probability theory to address the remaining service life and load sequence effects in a probabilistic manner. A comprehensive literature review, theoretical development of fatigue survival models for various engineering applications, and verification of these models using existing or new experiments, and synthesis of results constitute the scope of this research.