A B S T R A C T Due to the variables and unknowns in both material properties and predictive models in creep crack growth (CCG) rates, it is difficult to predict failure of a component precisely. A failure strain constraint based transient and steady state CCG model (called NSW) modified using probabilistic techniques, has been employed to predict CCG using uniaxial data as basic material property. In this paper the influence of scatter in the creep uniaxial properties, the parameter C * and creep crack initiation and growth rate have been examined using probabilistic methods. Using uniaxial and CCG properties of C-Mn steel at 360 • C, a method is developed which takes into account the scatter of the data and its sensitivity to the correlating parameters employed. It is shown that for an improved prediction method in components containing cracks the NSW crack growth model employed would benefit from a probabilistic analysis. This should be performed by considering the experimental scatter in failure strain, the creep stress index and in estimating the C * parameter.
Keywords Crack growth rate
N O M E N C L A T U R Ea = crack lengtḣ a = creep crack growtḣ a i = creep crack growth rate at ith elemenṫ a ss = creep crack growth rate under steady state A = creep coefficient in creep rate b = remaining ligament ahead of crack B n = net section thickness between side-grooves C * = steady state creep characterising parameter e n , e ν , e ε f , e φ , eε, e tr and eȧ = error in n, ν, ε f , φ,ε, t r andȧ, respectively F = geometry factor for experimental estimates of C * D = constant for creep crack growth correlation by C * h 1 = non-dimensional function of geometry H = coefficient of rupture law I n = amplitude function in HRR field K = stress intensity factor n = creep exponent P = applied load P lc = limit load r c = size of creep zone ahead of crack tip t r = time to ruptureCorrespondence: M. Yatomi.