Reducing uncertainty in fatigue life limits of turbine engine alloys

Abstract

a b s t r a c tIn probabilistic design of materials for fracture-critical components in modern military turbine engines, a typical maximum design target risk (DTR) is 5 Â 10 À8 component failures/engine flight hour. This metric underscores the essential role of safety in a design process that simultaneously strives to achieve performance, efficiency, reliability, and affordability throughout the life cycle of the engine. Traditionally, the design and life management approaches for engine materials have typically relied on extensive testing programs to produce large databases of fatigue data, from which statistically based life limits are derived by extrapolation from the mean fatigue behavior. However, we have found that the statistical behavior of fatigue lifetimes under a given test condition often exhibits a bimodal form, and that the trends in mean vs. minimum fatigue lifetime typically respond differently to loading or to microstructural variables. Under such circumstances, the underlying life-limiting mechanisms appear to exhibit a probabilistic microstructural hierarchy in fatigue resistance that is controlled by susceptibility of local microstructural neighborhoods to early damage and the growth of small cracks. These findings suggest that significant opportunities may exist for reductions in uncertainty in materials life-cycle prediction and management, if such hierarchies can be understood and controlled. This paper explores the potential implications of these findings, and a number of possible approaches are suggested for incorporating the insights of life-limiting fatigue into methods of integrated computational materials engineering (ICME) to support optimized life-cycle design of materials and components in turbine engines. Benefits of this approach appear to include substantial improvements in model accuracy, coupled with reduced requirements for materials testing, potentially leading to a significant reduction in the time and cost to develop, validate, transition, and implement new, more fatigue resistant alloys. Published by Elsevier Ltd.

Making the Case for a Model-Based Definition of Engineering Materials

“…With access to a model-based description of a material, a part serial number, and digital thread data on aircraft use, sustainment engineers would be able to rapidly and accurately assess the effects of damage or mission profile changes onto component health. Additionally, operational systems could serve as flying laboratories, providing crucial performance data back to materials and design engineers in validating lifing models, or providing new insights for improvements [28].…”

“…Feedback loops could reduce the need for rediscovery of issues and solutions along the entire process and provide new insight to validity of assumptions made earlier in the process. Data analytic techniques could allow new material compositions to be discovered more readily, while operation of systems could effectively become field laboratories used to refine component lifing models using complex and long-term data streams [28,47,48]. MBMDs are ideal platforms for continuous improvement and harvesting of production and operational data for increased knowledge.…”

Probability of Life‐Limiting Fatigue Failures in Titanium Alloys

“…Fatigue lifetime data tends to capture the mean fatigue behavior quite well, for example a stress vs. lifetime or the S-N plot provides a good depiction of the mean lifetime trend and, depending on the number of tests at each stress level, also gives a sense for the distribution about the mean. The distribution with respect to the mean behavior, however, may not extrapolate to the realistic minimum lifetime limit [1,2]. This is because; the mean and the minimum behaviors are controlled by different mechanisms and respond at different rates to change in stress level, microstructure, and other variables [1,2,3].…”

“…Fatigue life prediction of fracture critical parts is often dictated by the minimum rather than the mean fatigue behavior [ 1]. In order to obtain the minimum lifetime limit, a statistical distribution is often fit to the fatigue data, which is extrapolated to the desired probability level, for e.g., the 1 in 1000 probability of failure or the B0.1 lifetime, to determine the minimum lifetime [1].…”

“…In order to obtain the minimum lifetime limit, a statistical distribution is often fit to the fatigue data, which is extrapolated to the desired probability level, for e.g., the 1 in 1000 probability of failure or the B0.1 lifetime, to determine the minimum lifetime [1]. Fatigue lifetime data tends to capture the mean fatigue behavior quite well, for example a stress vs. lifetime or the S-N plot provides a good depiction of the mean lifetime trend and, depending on the number of tests at each stress level, also gives a sense for the distribution about the mean.…”