Naslov
Characterization of parameters influencing friction in the nanometric domain

Autori
Perčić, Marko

Vrsta, podvrsta i kategorija rada
Ocjenski radovi, doktorska disertacija

Fakultet
Tehnnički fakultet

Mjesto
Rijeka

Datum
17.02

Godina
2020

Stranica
208

Mentor
Zelenika, Saša ; Mezić, Igor

Ključne riječi
nanometric friction ; atomic force microscopy ; nanotribology of thin films ; experimental measurements ; friction modelling

Sažetak
Friction and wear are recognized as one of the most puzzling problems, not only in many engineering and manufacturing applications, but also in a fundamental scientific sense. In fact, friction is a nonlinear stochastic effect with a distinct time, position and temperature variability. While frictional phenomena on the macro- and meso-scales can be considered well described, and their characteristic features can be simulated via suitable models, as well as generally efficiently compensated by using proper control typologies, the study of friction, the parameters that influence its value and the respective models in the nanometric domain are still in an early stage, due to various experimental and modelling complexities. The research performed in the framework of the doctoral thesis provides a scientific contribution to the study of dry (unlubricated) friction by characterising the parameters influencing its value at the nanometric scale, and especially the dependence of friction on material properties, loading conditions, the velocity of motion, as well as temperature. The characterisation of the dependence of friction on the listed parameters is based on experimental measurements performed by employing a Scanning Probe Microscope (SPM). Due to the number and variety of the monitored influences, the number and type of measurements is determined by a state-of-the-art Design of Experiment (DoE) methodology by employing Voronoi tessellations. To obtain predictive models linking the process variables to the value of nanometric friction, the obtained measurement results are then validated numerically via a thorough comparative analysis of state-of-the-art machine learning methods. Despite the marked complexity of the analysed phenomena and the inherent dispersion of the measurements, the developed symbolic regression models, show, depending on the type of the sample, an excellent prediction accuracy between 72 and 91%.

Izvorni jezik
Engleski

Znanstvena područja
Strojarstvo, Temeljne tehničke znanosti, Interdisciplinarne tehničke znanosti

POVEZANOST RADA