Page 164 - Data Science Algorithms in a Week

148 Djordje Cica and Davorin Kramar

mostly used training algorithm is backpropagation (BP) algorithm. However, some
inherent problems frequently encountered in the use of this algorithm, such are risk of
being trapped in local minima and very slow convergence rate in training have initialized
development of bio-inspired based neural network models. The objective of this study
was to utilize two bio-inspired algorithm, namely GA and PSO, as a training methods of
ANN for predicting of cutting forces in turning of Inconel 718 assisted with high pressure
coolant. The results obtained from the GA-based and PSO-based ANN models were
compared with the most commonly used BP-based ANN for their performance. The
analysis reveals that training of ANN by using bio-inspired algorithms provides better
solutions in comparison to a conventional ANN.

Keywords: cutting forces, high-pressure cooling, neural networks, genetic algorithms,
particle swarm optimization

INTRODUCTION

High performance manufacturing is an inclusive term incorporating many existing
theories and approaches on productivity and waste reduction. In recent years, different
cooling techniques have been applied in order to increase productivity of the machining
process. Tremendous opportunities in terms of improving the overall process
performance are offered by the high pressure cooling (HPC) technique which aims at
upgrading conventional machining using high pressure fluid directed into the tool and
machined material. The high pressure coolant allows a better penetration of the fluid into
the workpiece-tool and chip-tool interfaces, which results in a better cooling effect,
reduction of friction and improving tool life (Diniz & Micaroni, 2007; Kramar & Kopac,
2009; Wertheim, Rotberg, & Ber, 1992). Furthermore, high pressure coolant reduce the
tool-chip contact length/area, improve chip control and reduce the consumption of cutting
fluid (Ezugwu & Bonney, 2004).
Due to their mechanical, thermal and chemical properties, nickel-based alloys are
among the most commonly used materials in aerospace and chemical industry, power
production, environmental protection, etc. However, nickel-based alloys are considered
as materials that are hard to machine. The poor thermal conductivity of these alloys raises
temperature at the tool-workpiece interface during conventional machining (Kramar,
Sekulić, Jurković, & Kopač, 2013). Thus, short cutting tool life and low productivity due
to the low permissible rates of metal removal are inevitable associated with the
machining of nickel-based alloys. Conventional cooling is not efficient to prevent
extreme thermal loading in the cutting zone, so the focus of recent studies is aimed at
reducing temperature in the cutting zone by applying different cooling techniques.
Among them, HPC technique is starting to be established as a method for substantial
increase of removal rate and productivity in the metal cutting industry.