Modelling the influence of clustered defects on HCF properties of Nibased superalloys - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Modelling the influence of clustered defects on HCF properties of Nibased superalloys

Lionel Marcin
Vincent Maurel
Henry Proudhon

Résumé

Casting defects like shrinkages and pores are very detrimental for material's High Cycle Fatigue (HCF) performance. It is known that defect's size and their location from free surface are the most important parameters to take into account regarding HCF properties. However, this conclusion is questionable in the case of clustered defects: specimen lifetime seems to also depend on other characteristics of the defect cluster. The distribution of defect size in an individual cluster is found to be dependent on clustered defects volume and thus influencing the ASTM grades of the material. The gradient associated to the same isolated defect is modified in the presence of cluster further modifying the fatigue predictions via non-local approaches like volumetric homogenization. However, this type of analysis requires a very large number of three-dimensional observations of casting defects dataset which is impractical and very expensive. Therefore, in this work, a Generative Adversarial Network (GAN) is integrated with Convolutional Neural Network (CNN) to generate synthetic defects (see figure 1.a) and are distributed in material space with the aid of spatial point pattern analysis of defects to generate synthetic microstructures that mirrors real specimens as shown in figure 1.b. Four Inconel 100 and nine Rene 125 specimens tomographed via X-Ray Computed Tomography have been used to study the spatial point pattern of defects and to train Deep Neural Networks (DNN). Spatial point pattern analysis helps to understand how the defects are distributed in material space whilst GANs generate synthetic defects which can be placed randomly respecting the statistics of real specimens. These N synthetic samples are submitted to numerical simulations to exploit the fatigue behavior via non-local approaches, see figure 2. Furthermore, some of the important features of defect clusters such as, defect volumes, cluster thickness, density of defects in a cluster etc., are controlled to analyze the influence on stress gradients in the synthetic microstructures.
Fichier principal
Vignette du fichier
FATIGUE2023_Arjun.pdf (299.25 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04346313 , version 1 (15-12-2023)

Identifiants

  • HAL Id : hal-04346313 , version 1

Citer

Arjun Kalkur Matpadi Raghavendra, Lionel Marcin, Vincent Maurel, Henry Proudhon. Modelling the influence of clustered defects on HCF properties of Nibased superalloys. 13th International Fatigue Congress, Nov 2023, Hiroshima, Japan. ⟨hal-04346313⟩
18 Consultations
7 Téléchargements

Partager

Gmail Facebook X LinkedIn More