Parametric probabilistic models for predicting creep remaining useful life
Résumé
In industry, materials are often used in service at high temperatures and under mechanical stress, due to the need to increase efficiency imposed by increased consumption. A progressive deformation, depending on time, of a material under tension or constant loading is called creep. This metallic phenomenon is of great industrial interest and a forecast of the life of mechanical components in creep is a critical phase of the initial stages of the design of equipment that operates at high temperatures. When recognizing this problem, a detailed understanding of creep is essential for the success of these projects, ensuring that components do not undergo excessive deformation which can lead to failure. With this objective, several parametric methods were created to quantify the creep deformation in high temperature applications. However, most of them use a deterministic approach and do not consider the dispersion of experimental creep data in their forecast. In this sense, this work adopted a parametric probabilistic approach to quantify the uncertainties associated with the parameters of parametric models for predicting creep remaining life, considering a small amount of experimental data available. Based on the statistical information extracted from the experimental data, a theoretical framework for quantifying uncertainty based on Monte Carlo simulations was applied to determine the probability distribution of the model parameters. With this, it is possible to evaluate the forecasting capacity of each model and define safe limits with known confidence levels.