Perspektywy obliczeń neuronowych w inżynierii łożyskowania

Przedyskutowano zastosowanie sztucznych sieci neuronowych do przewidywania zużycia oraz skomentowano rezultaty takich prób poczynione przez innych badaczy. Stwierdzono potrzebę i możliwość takiego zastosowania. Ukazano błędy popełnione przez innych badaczy.

---

Feasibility of artificial neural networks for wear prediction and some results obtained by other researchers are discussed in this paper. Analysis of known wear prediction methods in bearing systems (or more generally tribo - systems) reveals the need for continued research in the area. Usability of artificial neural networks over a wide range of applications encourages using them in preliminary stage of data analysis for approximation of wear as a function of parameters and variables describing the bearing system. Papers published on the specific subject are few, yet their presence in important journals (Tribology International, STLE Tribology Transactions) proves the existence of a general interest in the subject. The interdisciplinary character of the task of applying neural networks in tribology brings about a high risk of erroneous actions and misinterpretation of results. A well theory laden approach must be used to eliminate the risk and a possible discouragement to this promising area of research.