Naslov
Nanotribological characterization of an X39CrMo17- 1 steel thin-film via measurement-based machine learning methods

Autori
Perčić, Marko ; Zelenika, Saša ; Mezić, Igor

Vrsta, podvrsta i kategorija rada
Radovi u zbornicima skupova, cjeloviti rad (in extenso), znanstveni

Izvornik
Conference Proceedings of the 22nd International Conference of the European Society for Precision Engineering and Nanotechnology / Leach, R. K. ; Akrofi-Ayesu, A. ; Nisbet, C. ; Phillips, D. - Cranfield : European Society for Precision Engineering and Nanotechnology (EUSPEN), 2022, 51-54

ISBN
978-1-9989991-1-8

Skup
22nd International Conference of the European Society for Precision Engineering and Nanotechnology

Mjesto i datum
Ženeva, Švicarska, 30.05.2022. - 03.06.2022

Vrsta sudjelovanja
Plenarno

Vrsta recenzije
Međunarodna recenzija

Ključne riječi
nanotribology ; high-alloy steel ; thin-films ; machine learning

Sažetak
Frictional characteristics are an important factor in the design and exploitation of precision positioning systems and components. In fact, friction is a complex stochastic phenomenon depending on an intricate interplay between materials’ physical and chemical properties, as well as on the contact conditions influenced by the exerted normal forces, sliding velocities, contact area, temperature etc. Experimental measurements of the nanoscale friction force of an X39CrMo17-1 high-alloy steel thin film, deposited on a silicon wafer, are obtained in this work via elaborated lateral force microscopy (LFM) measurements performed on a scanning probe microscope using silicon-nitride microcantilever probes. The thin-film samples are synthesised by using pulsed laser deposition, which enables obtaining precise stochiometric elemental properties. The LFM measurements allow obtaining single-asperity contact conditions in the studied tribo-pair, which is studied experimentally considering the mentioned external influences as the variable process parameters. In order to obtain an in-depth insight into the dependencies of nanoscale friction on these parameters, the obtained experimental data is used to train various machine learning (ML) algorithms. The resulting predicted values of the nanoscale friction force are therefore obtained by using random forest, multilayer perceptron and support vector regression ML methods. In order to assess their predictive performances, all the used algorithms are validated on a separately measured test dataset. The best predictive performances are obtained by using the support vector regression algorithm, resulting in the highest achieved coefficient of determination (R2), allowing the prediction of 90 % of the variance of the experimental measurements.

Izvorni jezik
Engleski

Znanstvena područja
Strojarstvo, Temeljne tehničke znanosti, Interdisciplinarne tehničke znanosti

POVEZANOST RADA