Fatigue life prediction at mesoscopic scale of samples containing casting defects: A novel energy based non-local model

TL;DR

This paper introduces a novel energy-based non-local model for predicting fatigue life at the mesoscopic scale in samples with casting defects, validated through experiments and detailed finite element simulations.

Contribution

The study develops a new non-local fatigue model that explicitly accounts for defect shape and stressed volumes, improving life prediction accuracy for cast nickel-based superalloys.

Findings

Model predicts fatigue life within a factor of 3 of experiments.

Finite element models reveal high stress concentrations in defect ligaments.

The approach generalizes non-local methods to real casting defects.

Abstract

Fatigue failure driven by stress gradients associated to casting defects was studied in two cast nickel-based superalloys. The experimental campaign revealed complex damage phenomena linked to spongeous shrinkages, characterized by their intricate arrangement of defects in the material medium, forming defect clusters. Multiple cracks were observed to initiate from defect volumes, coalescing with neighboring void surfaces along crystallographic planes. Defects were characterized using X-ray computed tomography, and image-based finite element (FE) models were constructed as digital representations of each experimental sample explicitly containing all real casting defects. Numerical simulations of these FE models under the same conditions as the experiments revealed that tortuous defects contain small ligaments where very high local stresses develop. These ligaments initiate early cracks, but due to the limited stressed volumes, these cracks do not drive the life of the material. A thorough comparison of simulations with experiments led to the development of an original method to define stressed volumes and address small ligaments. Finally, a novel energy-based non-local model was proposed, using two parameters to predict the fatigue lives of samples containing casting defects at the mesoscopic scale. The model was validated against samples with varying porosity levels, sizes, and geometries, accurately predicting fatigue lives within a factor of 3 compared to experimental results. This new approach generalizes the application of non-local methods to real casting defects by considering their shape and stressed volumes to estimate fatigue properties.