| Summary: | Fatigue assessments have been carried out predominantly with quasi-deterministic approaches, such as the use of SN curves. However, both the loading and the resistance of fatigue prone components are subjected to significant uncertainties. Consequently, a prediction of the remaining fatigue life based on deterministic load and resistance models can lead to unreliable results. This work presents a general reliability-based approach to predict fatigue life of steel components. The approach incorporates prediction of fatigue crack initiation, modeled with a strain-based correlation approach, and propagation, modeled using a linear elastic fracture mechanics approach, and is applicable to new, cracked or repaired structural components.
Based on the analysis of existing test results and additional crack initiation and propagation tests on weld metal, the relevant probabilistic fatigue material properties of grade 350WT steel and a matching weld metal were established. An experimental program was carried out on welded details tested either in the as-welded, stress-relieved, conventionally peened, or ultrasonically peened condition. It was demonstrated that ultrasonic peening is superior to the other investigated post weld treatment methods. Using finite element analyses, the results of the tests were deterministically predicted for several different initial conditions, including initial flaw and crack sizes and locations, as well as different levels of residual stresses. A model incorporating an initial flaw and accounting for crack closure and the threshold stress intensity factor range was retained.
A probabilistic analysis using Monte Carlo Simulation was carried out to calibrate the relevant parameters. A general reliability-based approach, which includes both the loading and resistance sides of the limit state function was proposed and applied to three practical examples: prediction of test results from two test programs and the prediction of the remaining fatigue life of a cracked component as a function of the safety index. These three applications demonstrated that accurate fatigue life predictions targeting a predefined safety index are achieved. === Structural Engineering
|
|---|
