Effect on Machinability Characteristics of Cryogenic Process and Performance Assessment by Using Machine Learning Approach with Scaled Conjugate Gradient Algorithm

Abstract

The present investigation is focused on the machinability and characterisation of cryogenised AISI 4140 steels, which are renowned for exhibiting superior hardness, durability, wear resistance, dimensional and chemical stability, and fatigue strength when compared to conventional steels. Hard turning experiments were conducted on cryogenised steel specimens employing dry cutting conditions with a carbide insert. As the cutting tip, the TT5100 quality TIN coated type with code WNMG 080408 MT produced by TaeguTec company was used. The study examines the impact of various cutting parameters, including three distinct cutting speeds (160, 200, 240 m/min), three feed rates (0.04, 0.08, 0.12 mm/rev), and three depths of cut (0.1, 0.15, 0.2 mm), on power consumption and surface roughness values. The utilisation of the artificial neural network (ANN) approach, a machine learning methodology, for the analysis of measured values adds another layer of originality to the research. Transfer functions, such as Scaled Conjugate Gradient (Trainscg), were employed within the ANN structure. The calculated metrics include a mean absolute error (MAE) of 0.0218, 0.0542, and 0.0064, and a mean squared error (MSE) of 0.0429. The mean absolute error (MAE) was 0.0683, the mean squared error (MSE) was 0.0429, and the mean relative deviation (ARD%) was 10.71%, 0.0718%, and 0.2020%. Furthermore, the optimal values for the correlation coefficient (R2) were determined as 0.9512, 0.9999, and 0.9997.