A Probabilistic Creep Fatigue Computer Program for Structural Integrity Analysis

This paper describes a computer program which can determine the probability of failure of gas turbine structures as a function of time due to creep fatigue crack growth. The probability of failure is computed by combining stress analysis and creep fatigue analysis with probabilistic analysis methods. The creep fatigue analysis is based on a reference stress approach which provides a simple, accurate, and efficient method for determining the steady state component, C*, of the time dependent fracture mechanics parameter C(t). Stress intensity factors are computed from stress distributions derived from a linear elastic finite element analysis of the uncracked structure and weight functions. Several probabilistic methods are available such as efficient approximate methods, importance sampling and Monte Carlo sampling. Efficient approximate methods and importance sampling methods are typically one to two orders of magnitude more efficient than Monte Carlo sampling. Probabilistic sensitivity measures are generated as a byproduct of the probabilistic analysis and indicate the importance of the random variables to the reliability of the structure. The theoretical background, computer code and an example problem are presented.