Abstract
In this study, the effect of kinetic energy of the shot peening process on microstructure, mechanical properties, residual stress, fatigue behavior and residual stress relaxation under fatigue loading of AISI 316L stainless steel were investigated to figure out the mechanisms of fatigue crack initiation and failure. Varieties of experiments were applied to obtain the results including microstructural observations, measurements of hardness, roughness, induced residual stress and residual stress relaxation as well as axial fatigue test. Then deep learning approach through neural networks was used for modelling of mechanical properties and fatigue behavior of shot peened material. Comprehensive parametric analyses were performed to survey the effects of different key parameters. Afterward, according to the results of neural network analysis, further experiments were performed to optimize and experimentally validate the desirable parameters. Based on the obtained results the favorable range of shot peening coverage regarding improved mechanical properties and fatigue behavior was identified as no more than 1750% considering Almen intensity of 21 A (0.001 inch).Graphic abstract
Highlights
Fatigue failure mostly initiates from the surface layer of the components [1,2,3]
In order to obtain the structure of neural networks (NN) with highest performance and in order to compare the efficiency of Shallow neural network (SNN), deep neural network (DNN) and SAE assigned DNN (SADNN) various networks with different architecture and network parameters were developed for each considered modelling of A, B and C
Gradient structures from nano-structured to refined grains were generated in the surface layer by increasing the severity of shot peening compared to the conventional treatment
Summary
Fatigue failure mostly initiates from the surface layer of the components [1,2,3]. applying severe plastic deformation (SPD) methods such as shot peening (SP) have critical role for surface and structural integrity of the materials [2, 5,6,7,8,9]. Beneficial effects of conventional shot peening (CSP) and severe shot peening (SSP) which has higher severity than CSP (by increasing the values of intensity and coverage) were studied on improvement of mechanical properties and fatigue behavior of the different metallic materials in the last decade. SNN has 1 or 2 hidden layers which generally trained by back-propagation (BP) algorithm [40] These networks besides their beneficial applications have some limitations. After further improvements in this area, other alternative methods for pre-training of DNN such as stacked auto-encoder (SAE) were presented which helps to develop DNN with small data set and achieve higher efficiency by increasing the number of hidden layers and using SAE in between them [44,45,46,47]. Afterward, based on the obtained results of NNs, further experiments were accomplished to optimize and experimental validation of the desirable parameters as well as the specification of the boundary between the SSP and OSP
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
