Modeling of surface roughness in wire electrical discharge machining of Inconel 718 using artificial neural network. (2021)

• Record Type: Journal Article
• Authors: Paturi, Uma Maheshwera Reddy; Devarasetti, Harish; Reddy, N.S.; Kotkunde, Nitin; Patle, B.K.
• Is Part Of: Materials today. Volume 38:Part 5(2021)
• Page Start: 3142
• Page End: 3148
• Publication Date: 2021
• Subjects: WEDM -- Inconel 718 -- Surface roughness -- ANN -- Process modeling; Materials science -- Congresses -- Periodicals; 620.1
• DOI: 10.1016/j.matpr.2020.09.503
• Languages: English
• ISSNs: 2214-7853

Abstract

Abstract: Application of artificial neural network (ANN) in process modelling and parameter optimization has become quite obvious because of its capability to predict the output quickly and precisely. The current study attempts to model and predict the surface roughness in wire electrical discharge machining (WEDM) of Inconel 718 using artificial neural network (ANN). A multilayer perception model with back-propagation neural network (BPNN) is utilized to model the process. WEDM experimental data of this study has been divided into training, testing and validation data groups in the ratio of 5:1:1. Hyperbolic tangent sigmoid ( tansig ) and Levenberg Marquadt ( TrainLM ) were considered as the transfer function and training function respectively. ANN model comprises 5 neurons (peak current, voltage, pulse on time, pulse off time and wire electrode feed rate) in the input layer and 1 neuron (surface roughness) in the output layer. The performance indices considered were mean squared error (MSE) and average absolute error in prediction (AEP). The obtained optimal ANN structure comprises five neurons in input layer, eleven neurons in hidden layer and one neuron in the output layer (5-11-1). ANN outcome was then related with the experimentally acquired data. The ANN predictions were found to be in very highly agreement with experimentally measured results and yielding a correlation coefficient (R-value) as high as 99.8%. The outcome demonstrates that the ANN method is the efficient tool for the parameter optimization in WEDM process.