Numerical Stability of Creep Models for Metallic Materials in Large Turbochargers

Sammanfattning

This thesis investigates the numerical behavior, stability, and robustness of 7 different creep material subroutines used in finite element simulations of high-temperature components, with a focus on materials used on the turbine-side casings in large turbochargers. The aim is to develop a structured approach to evaluate existing creep subroutines and to identify factors governing their numerical performance. A numerical test suite was developed to systematically assess the industrial creep subroutines, used by Accelleron, which are associated with different cast iron materials. The methodology combines single-element simulations, enabling isolation of material behavior, with multi-element simulations that introduce geometrical complexity, stress redistribution, and temperature gradients. In addition, the contribution of individual terms within the underlying creep formulations was analyzed. Simulations were carried out in Abaqus, with post-processing performed in Matlab. The results show that the investigated subroutines differ significantly in numerical stability and predictive accuracy. Only a few of the models reproduced experimental creep behavior with low error, while others exhibited deviations of up to approximately 50%. Stability was generally maintained within the calibrated stress–temperature range but instability started showing under increased loading, particularly at elevated temperatures. An important distinction is that diverged solutions still produced relevant results up to 2% creep strain, but the solution diverged at a later stage. Differences in numerical behavior were found to be governed by the creep constants to a large extent. However, the constants act through the mathematical structure of each creep formulation, such that stability is controlled by the interaction between formulation and calibration rather than by either of them alone. It was concluded that a structured numerical test suite is a working method for evaluating creep subroutines. It enables the possible identification of stability limitations, parameter sensitivity, and model robustness under industrially relevant conditions. Although some models exhibit instability at extreme conditions, most produce physically meaningful results within engineering-relevant strain levels, highlighting the distinction between practical usability and numerical robustness.