Abstract
Creep data are often analyzed using derived engineering parameters to correlate creep life (either time to rupture, or time to a specified strain) to applied stress and temperature. Commonly used formulations include Larson-Miller, Orr-Sherby-Dorn, Manson-Haferd, and Manson-Succop parameterizations. In this paper, it is shown that these parameterizations are all special cases of a common general framework based on a linear statistical model. Recognition of this fact allows for statistically efficient estimation of material model parameters and quantitative statistical comparisons among the various parameterizations in terms of their ability to fit a material database, including assessment of a stress-temperature interaction in creep behavior. This provides a rational basis for choosing the best parameterization to describe a particular material. Furthermore, using the technique of maximum likelihood estimation to estimate model parameters allows for a statistically proper treatment of runouts in a test database via censored data analysis methods, and for construction of probabilistically interpretable upper and lower bounds on creep rate. A generalized Larson-Miller formulation is developed, which is comparable in complexity to the Manson-Haferd parameter, but utilizes a reciprocal temperature dependence. The general framework for analysis of creep data is illustrated with analysis of Alloy 617 and Alloy 230 test data.