Preliminary Diagnostics of Dynamic Systems From Time Series

Abstract

The state space reconstruction technique was recognized by Edward N. Lorenz as “one of the most surprising developments in nonlinear dynamics” [1]. Nowadays, the technique is applied in various scientific areas for prediction, analysis and diagnostics. This paper aims to discuss the possibility of using the embedding dimension of a reconstructed state space of time series as a tool for preliminary diagnostics. After a short description and illustration of the method, the paper considers two case studies: a single degree of freedom (DOF) and a 2 DOF system. The results of the analysis help detect a class of structural defects, including defects connected to a coupling mechanism. There is clearly a huge potential of such an approach for diagnostics of complex machinery.