Modelling the corrosion behaviour of laser-deposited functionally gradient materials by experimentation and Bayesian regularisation algorithm

This study employed a Bayesian Regulation-back propagation neural network (BR-BPNN) model to estimate the electrochemical behaviour of laser deposited functionally gradient material (FGM) samples made from SS-316L and Co-Cr-Mo alloy, where the samples are laser deposited using process parameters as...

Full description

Saved in:
Bibliographic Details
Published inNeuroQuantology Vol. 20; no. 10; p. 4730
Main Authors Yakkaluri Pratapa Reddy, Kavuluru, Lakshmi Narayana, Mantrala Kedar Mallik
Format Journal Article
LanguageEnglish
Published Bornova Izmir NeuroQuantology 01.01.2022
Subjects
Algorithms
Artificial neural networks
Back propagation networks
Bayesian analysis
Chromium
Cobalt base alloys
Electrochemical analysis
Experimentation
Feed rate
Functionally gradient materials
Impedance
Laser deposition
Lasers
Low frequencies
Machine learning
Mathematical models
Molybdenum
Neural networks
Process parameters
Regularization
Surface roughness
Online AccessGet full text

Cover

More Information
Summary:This study employed a Bayesian Regulation-back propagation neural network (BR-BPNN) model to estimate the electrochemical behaviour of laser deposited functionally gradient material (FGM) samples made from SS-316L and Co-Cr-Mo alloy, where the samples are laser deposited using process parameters as laser power, (LP); powder feed rate, (PFR); and scan velocity, (SV). The output data obtained by experimentation is considered input data to carry out the BR-BPNN analysis. Of the total experimental data, 70% wastaken to train the neural network, 15% for testing and 15% to validate the network. A high scan velocity (SV) with elevated laser power (LP), added with mild powder feed rate (PFR), exhibits a Cole-Cole (Nyquist) plot with a low-frequency impedance at a high-frequency zone. A Warburg spike was observed in the low-frequency band for all the samples, with variations in phase shift angle (ϕ) brought about by the influence of the process parameters. A depressed semicircle was observed in the high-frequency region for the samples deposited with high laser power. The impedance results of both experimental and machine learning approaches were compared. The FGM samples were subjected toa surface roughness test before an impedance test and found an average surface roughness (Ra) range of 2.25 to 2.40 µm.
ISSN:1303-5150
DOI:10.14704/nq.2022.20.10.NQ55453