Enhancing Cutting Oil Efficiency with Nanoparticle Additives: A Gaussian Process Regression Approach to Viscosity and Cost Optimization

TL;DR

This study uses nanoparticle additives and Gaussian process regression to optimize cutting oil efficiency and cost in machining.

Contribution

A novel fitness function combining viscosity and cost is proposed for optimal nanofluid selection.

Findings

ZnO and ZnO hybrid mixtures showed the best performance in terms of viscosity and cost.

GPR estimates enabled accurate prediction of dynamic viscosity with R2 = 1.

Limited experiments combined with GPR can guide optimal nanofluid selection without full testing.

Abstract

Nanoparticle additives are used to increase the cooling efficiency of cutting fluids in machining. In this study, changing dynamic viscosity values depending on the addition of nanoparticles to cutting oils was investigated. Mono nanofluids were prepared by adding hBN (hexagonal boron nitride), ZnO, MWCNT (multi-walled carbon nanotube), TiO2, and Al2O3 as nanoparticles, hybrid nanofluids were prepared by using two types of nanoparticles (ZnO + MWCNT, hBN + MWCNT etc.), and ternary nanofluids were prepared by using three types of nanoparticles. GPR (Gaussian process regression) was used to estimate unmeasured dynamic viscosity values using the dynamic viscosity values measured for different temperatures. Dynamic viscosity results are a precise determination (R2 = 1). An augmented dataset was obtained by adding the dynamic viscosity values estimated with high accuracy. A fitness function based on dynamic viscosity and nanoparticle unit costs was proposed for the cost analysis. With the help of the proposed fitness function, it was observed that the best performing nanoparticles were the ZnO and ZnO hybrid mixtures according to different dynamic viscosity and cost effects. The study showed that the most suitable nanofluid selection focused on performance and cost could be made without performing experiments under various operating conditions by increasing the limited experimental measurements with strong GPR estimates and using the proposed fitness function.

Figures