Modelling of Rotary EDM Process Parameters of Inconel 718 Using Artificial Neural Networks

Jayaraj Jeevamalar,Chinnamuthu Senthilkumar,Sundaresan Ramabalan

doi:10.5755/j01.mech.26.6.20484

Jayaraj Jeevamalar, Chinnamuthu Senthilkumar + Show 1 more

Open Access

https://doi.org/10.5755/j01.mech.26.6.20484

Copy DOI

Abstract

Modelling is used for correlating the relationship between the input process parameters and the output responses during the machining process. To characterize real-world systems of considerable complexity, an Artificial Neural Network (ANN) model is regularly used to replace the mathematical approximation of the relationship. This paper explains the methodological procedure and the outcome of the ANN modeling process for Electrical Discharge Drilling of Inconel 718 superalloy and hollow tubular copper as tool electrode. The most important process parameters in this work are peak current, pulse on time and pulse off time with machining performances of material removal rate and surface roughness. The experiments were performed by L20 Orthogonal Array. In such conditions, an Artificial Neural Network model is developed using MATLAB programming on the Feed Forward Back Propagation technique was used to predict the responses. The experimental data were separated into three parts to train, test the network and validate the model. The developed model has been confirmed experimentally for training and testing in considering the number of iterations and mean square error convergence criteria. The developed model results are to approximate the responses fairly exactly. The model has the mean correlation coefficient of 0.96558. Results revealed that the proposed model can be used for the prediction of the complex EDM drilling process.

Highlights

Nowadays the researchers utilize Artificial Neural Networks (ANN) for demonstrating complex modern industrial issues
Fenggou & Dayong [12] proposed an ANN modelling technique to find the number of hidden nodes and optimize the correlation between input variables and performance measures using Genetic Algorithm (GA) and Back Propagation Learning Algorithm (BPLA)
Several researchers reported that the Back Propagation algorithm is the most appropriate approach for handling large learning problems

Summary

Introduction

Nowadays the researchers utilize Artificial Neural Networks (ANN) for demonstrating complex modern industrial issues. Fenggou & Dayong [12] proposed an ANN modelling technique to find the number of hidden nodes and optimize the correlation between input variables and performance measures using Genetic Algorithm (GA) and Back Propagation Learning Algorithm (BPLA). This system has been examined for the optimization of multiple parameters (e.g. SR, Micro-Hardness (MH), the thickness of the recast layer and MRR) of the process In this approach, R-square and Means Square Error (MSE) are used to evaluate the efficiency of ANNs. Several researchers reported that the Back Propagation algorithm is the most appropriate approach for handling large learning problems. The Back Propagation Neural Networks (BPNN) scheme has mathematically strong learning ability in training and relating input and output variables [15] They are usually called feed-forwarded, multi-layered networks. Hidden layers are used to perform nonlinear transformations on the input space and are used for computation purposes

Design of experts

ANN model description

PURELIN TRAINLM LEARNGDM

Verification of trained networks

Summary