Prediction of Structural Crack Growth Behavior under Fatigue Loading

Abstract

Structural components which undergo operational fatigue cycles have historically been designed for fatigue resistance based upon crack initiation data generated on smooth and notched laboratory speimens. More recently, reliability has been improved by using linear elastic fracture mechanics (LEFM) to assure that flaws, even if present, cannot grow to a dangerous size during service. The methodology and input data utilized for LEFM evaluation are described, and the impact of variations in key input parameters are discussed in detail. Specifically, the paper describes the sensitivity to (1) loading parameters: applied cyclic stresses, mean stresses, residual stresses, load spectrum, and stress distributions, (2) local crack driving force: crack size, shape, orientation, growth, and growth mode, (3) materials properties: crack growth rate constants, crack growth threshold, and response to spectrum overloads, and (4) initial flaw size, with and without nondestructive inspection. Presently, LEFM analyses are often deterministic and utilize conservative (worst-case) assumptions for all input parameters which are not known. The assumptions that all worst-case conditions occur simultaneously lead to extremely conservative designs which may be economically unacceptable. Probabilistic fracture mechanics methods have been developed which incorporate the statistical variations or uncertainties that actually exist and establish a more realistic quantitative basis for design allowables and accept/reject criteria, which optimize the cost/risk trade-offs that must be made.