Four-Stage Model for Predicting Creep Behavior

This technical note presents the computational foundations of an algorithm used to predict the creep behavior of material under stress. Given time-strain pairs of test data, the algorithm finds a best-fitting curve from a class of nonlinear models by using variable reduction and numerical searches. The numerical searches are proven to be monotonically convergent. The model class employs four components with matched derivatives to represent the traditional primary, secondary, and tertiary stages of creep. Algorithm output is a collection of parameters describing the curve from the model class that best represents strain as a function of time under constant stress and temperature. Since the output parameters characterize creep rate and creep stage demarcations, multiple samples of the same material under different stresses and temperatures can be used to model material creep rate as a function of temperature and/or stress. Several example fits to real metallurgical test data are provided.