Creep crack growth assessments are generally performed using a number of in-house procedures and/or national standards (e.g. A16, BS7910 and R5). These methods are based on a deterministic approach and depending on the assumptions made the material properties employed, significant variations in predictions can be obtained. However, probabilistic approaches and sensitivity analyses can be carried out to determine the most important variables influencing the predictions. In this paper, Monte Carlo Simulations (MCS) and sensitivity analyses are performed on creep crack growth data generated on a 9%-Cr steel and 2(1/4)Cr1Mo ferritic steel designated P91 and P22, respectively, using the creep fracture mechanics parameter C~* and the Norton creep law for describing the creep strain rate material properties. Three aspects have been investigated. Firstly, it is demonstrated that the MCS method is capable of reproducing the "cloud" of experimental scatter data accurately. Secondly, the number of simulations ' required to establish satisfactory predictions at low failure probabilities (< 1%) is examined. Finally, sensitivity analyses have been undertaken to determine the range in predictions that can be obtained from the statistical variations in material properties measured. The most important parameters and variations are identified for each steel and general advices are given.
展开▼
