A multiscale approach for prediction of failure probabilities of engine components

The present work focuses on a complete procedure to compute the failure probabilities of different engine components using multiscale simulations taking microstructural characteristics into account and Weibull's Weakest Link Theory. The Weakest Link Theory evaluates the failure probability on the basis of experimental measurements. Multiple tests are required to obtain the scatter of the required fracture stress, which are limited by high costs of manufacturing processes and measurement techniques. To circumvent this issue experimental results can be complemented by multiscale simulations. With the application of a homogenization process the overall material modulus is determined based on microscopic properties, such that the morphology of the microstructure, defects or pores can be directly incorporated. In this paper we use the FE2-method with the idea to assign a representative volume element (RVE) to each macroscopic integration point, instead of deriving a suitable macroscopic material model. This RVE reflects the properties of a realistic heterogeneous microstructure and represents the overall material behavior. The propagation of micro cracks is simulated using a phase field model and is implicitly included in the homogenization process. Multiple variations of RVEs with small geometrical differences capture variations in the manufacturing process and result into a scattering of the fracture stress. With the obtained results a subsequent Weibull analysis can be performed, resulting into a prediction of the failure probability of different engine components.