A Data Assisted Prediction of Fatigue Life for Aluminum Alloys Using
Machine Learning Approach
The article presents a machine learning (ML) model capable of
recognizing the effect of physical and chemical parameters that
contribute to fatigue failure in aluminium alloys. The traditional
method of obtaining S-N curve is both expensive and time-consuming. The
mechanism of fatigue is complex and influenced by a number of factors.
The current study proposes a data-driven method to estimate fatigue life
at different stress amplitudes that form the S-N curves. The influencing
factors dominating the fatigue life can be effectively integrated using
Machine Learning techniques to predict S-N curves of aluminium alloys.
Dataset was prepared from industrially accepted references. MLP and GBR
algorithms were employed to train the model. The prediction of fatigue
life had a MSE of 0.46. It is interesting to note that the prepared
model could recognize the features that most affected the fatigue life
and predict the S-N curve which had close agreement with the
experimental data. The current study intends to assist material
scientists and design engineers to investigate the influence of
different alloying element compositions on fatigue life. The model can
be employed to obtain a preliminary estimate of fatigue life resulting
from varying alloy mixtures.