Original Article

Artificial neural network-based modelling and prediction of white layer formation during hard turning of steels

¹ Laboratory of Mechanics, Materials and Processes (LMMP), National High School of Engineering of Tunis (ENSIT)−University of Tunis, 5 Avenue Taha Hussein, Montfleury 1008 Tunis, Tunisia
² Department of Mathematics and Industrial Engineering, Polytechnic Montreal, Montreal QC H3T 1J4, Canada
³ Sousse Higher Institute of Applied Sciences and Technology (ISSATSo)−University of Sousse, Cité Taffala, Ibn Khaldoun 4003 Sousse, Tunisia

^* e-mail: souidabdallah123@gmail.com

Received: 30 August 2023
Accepted: 4 January 2024

Abstract

During hard machining, steels subjected to very high thermal and mechanical loads can result in microstructural/phase changes such as the formation of a white layer. This layer, which is often harder than the raw material, is considered detrimental to the fatigue performance and in-service life of machined parts. This paper proposes a comprehensive study of white layer formation during hard machining of steels using statistical analysis and artificial neural networks (ANN) modeling. To this end, two steals, named AISI 52100 and AISI 4340, commonly used in the manufacturing of structural machines’ components and extensively studied in the last decade, have been considered in this study. First, Taguchi method combined with response surface methodology (RSM) was applied to analyze and to optimize the machining parameters regarding the white layer thickness. Second, an ANN model is developed to predict the white layer thickness during hard machining of the studied steels using a large amount of machining data. Three training algorithms were tested to find the most robust configuration. The equivalent carbon parameter was introduced for the first time in machining modeling which make the proposed ANN-based model capable of predicting the white layer thickness for different hardened steels. The results show a significant agreement between predictions and experimental results, avoiding costly experimental machining tests.