Inférence statistique de la population de défauts pour l'étude et la simulation de la fatigue oligocyclique d'un alliage d'aluminium de fonderie

Abstract : In a search for lightweight materials, cast aluminium alloys have become an efficient solution for the automotive industry. For cylinder heads, cast aluminium alloys (Aluminium-Silicon-Copper based) are often chosen for their mechanical performances combined with their good casting abilities. This study aims at better understanding the damage mechanisms of Low-Cycle Fatigue (LCF) to help define a suitable design criterion. This is achieved by combining experiments (LCF tests at ambient and high temperature), various observations (Scanning electron microscopy, laboratory 3D X-ray microtomography, crack propagation monitoring) and numerical techniques (Finite element method simulations and defect population generation). Specifically designed statistical tools allow clearly identifying the defect (casting defects) population: the defect size can be modelled by a generalized Pareto distribution and their positions by a clustered point process (thus showing the defect locations are not completely random). Using marked point processes, it is also shown that defect size and defect position are uncorrelated. Using this statistical information and by systematic observations, it is proven that only defects close to the surface are critical for the fatigue life. The different interactions between cracks and defects are also studied, especially with respect to temperature. A strain energy based fatigue criterion is introduced and allows estimating the fatigue life. Then, a crack propagation model provides a mean of taking the defect size into account, which combined with the statistics of defects, gives an estimate of the fatigue life dispersion. Finally, notched specimens tests and simulations show the study of the defect population is even more critical as the highly loaded volume becomes smaller.
Complete list of metadatas

https://pastel.archives-ouvertes.fr/tel-01511338
Contributor : Abes Star <>
Submitted on : Thursday, April 20, 2017 - 5:36:10 PM
Last modification on : Friday, February 22, 2019 - 1:28:03 AM
Long-term archiving on : Friday, July 21, 2017 - 2:04:12 PM

File

WILSON.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01511338, version 1

Collections

Citation

Pablo Wilson. Inférence statistique de la population de défauts pour l'étude et la simulation de la fatigue oligocyclique d'un alliage d'aluminium de fonderie. Mécanique des matériaux [physics.class-ph]. Ecole nationale supérieure d'arts et métiers - ENSAM, 2017. Français. ⟨NNT : 2017ENAM0001⟩. ⟨tel-01511338⟩

Share

Metrics

Record views

854

Files downloads

310