Vol.6, Special Issue, 2007 pp. 185-196
UDC 007.52:004.896(045)=111

PARAMETER ESTIMATION FOR ONLINE CONDITION MONITORING OF ROBOTIC MACHINES
Tsz Ming James Hui, Honghai Liu, David J. Brown
Institute of Industrial Research, University of Portsmouth Burnaby Building, Portsmouth PO1 3QL, England, United Kingdom
e-mail: Honghai.Liu@port.ac.uk; David.Brown@port.ac.uk

Abstract. This paper proposes a novel learning approach to online condition monitoring of robotic machines. The real-time learning process comprises three stages, domain knowledge defining, random learning and ordinal learning. Domain knowledge defining abstracts the model of a robotic machine; random learning and ordinal learning stages train the parameters of the abstract model with random data selection and ordinal data selection, respectively. Simulation results have proved that the pro-posed method is efficient and feasible for online fault diagnosis of robotic machines.
Key words: Online machine learning, fault diagnosis and machine monitoring

PROCENA PARAMETARA ONLAJN USLOVA MONITORINGA ROBOTSKIH MAŠINA
Ovaj rad predlaže novi pristup u onlajn učenju uslova monitoringa robotskih mašina. Realno vreme procesa učenja sastoji se iz tri faze, domena odredjivanja znanja, slučajnog učenja i podregjenog učenja. Odredjivanje domena znanja sažima model robotske mašine; faze slučajnog učenja i podregjenog učenja utiču na parametre sažetog modela sa slučajno odabranim podacima i podregjenog odabranim podacima ponaosob. Rezultati simuliranja dokazuju da je predloženi metod efikasan iI izvodljiv za onlajn dijagnozu grešaka robotskih mašina.
Ključne reči: Onlajn učenje, mašinsko učenje, dijagnoza grešaka, monitoring robotskih mašina