An improved Artificial Neural Network for the direct prediction of fretting fatigue crack initiation lifetime

Fretting fatigue is a common type of problem in the aviation and other engineering fields. Due to its multiaxial characteristics, it leads to a shorter overall fatigue life compared to plain fatigue conditions. Fretting fatigue crack initiation lifetime is a crucial part of the total lifetime, and the currently dominant method for research on fatigue behavior is the combination of the theoretical and numerical models. With the advent of the era of data science, machine learning has been widely used to predict fatigue behavior, but there are no many applications in the field of fretting fatigue. This paper proposed an improved Artificial Neural Network (ANN) using Balancing Composite Motion Optimization (BCMO) for quick prediction of fretting fatigue crack initiation lifetime. Physical-mechanical reasoning parameters, axial stress amplitude, shear stress amplitude, half contact width, and half stick zone width are considered as input parameters, and fretting fatigue crack initiation lifetime is set as the output feature. The main aim of BCMO is to improve the robustness of the ANN based on the influential pa-rameters, namely bias and weight. The provided results using ANN-BCMO are robust compared to ANN and traditional techniques from the literature. The Matlab code of improved ANN using BCMO can be found at https ://github.com/Samir-Khatir/BCMO-ANN.git

Automated summary

Fretting fatigue, a common problem in aviation and engineering fields, has a shorter fatigue life compared to plain fatigue due to its multiaxial characteristics. This paper proposes an improved Artificial Neural Network (ANN) using Balancing Composite Motion Optimization (BCMO) to predict fretting fatigue crack initiation lifetime quickly.