Non-linear dynamics tools for the motion analysis and condition monitoring of robot joints

Time series from non-damaged and three types of damaged robot joints are considered and analysed from the viewpoint of non-linear dynamics. The embedding spaces for the four types of signals are recovered. The application of surrogate data tests is used to prove the presence of non-linearities in the joints. The results suggest a rise in unstable behaviour due to the introduction of backlash in robot joints. The chaotic behaviour gets stronger with the increase of the backlash extent. This is confirmed by the increase of the embedding dimension as well as by the increase of the Lyapunov exponents and the correlation dimension with the backlash increase. A straightforward method for condition monitoring using non-linear dynamics characteristics, based on a classification procedure, is suggested. (C) 2001 Academic Press.