Multidimensional damage state identification using phase space warping

Abstract

A concept called phase space warping (PSW) has been introduced to describe small drifts (caused by damage) in the phase space of a system. In previous research, a multidimensional damage identification approach has been developed based on this concept and validated using numerical model simulation. Further experimental validation and extension will be main focus of this dissertation. At first, an updated version of a well-known two-well magneto-elastic oscillator is used to validate the damage identification approach experimentally. Good identification results are observed when fatigue damage and simulated damage are present independently or combined with each other. Then, discussion about how to improve performance by selecting parameters is given. Although optimal parameters are still not available, several basic selection rules are achieved. After that, a practical extension of the PSW-based identification approach is introduced called local flow variation (LFV), which is based on the probability distribution functions of trajectories in the phase space. Although the LFV-based approach cannot achieve identification results as good as the PSW-based one, it decreases computational time by two orders of magnitude. Finally, damage prognosis is realized based on identification results from the PSW-based approach using least square method. Several future research topics are proposed, including identifying damage based on multivariate signals, qualifying severity of damage, orthogonalizing damage tracking modes, and applying the LFV-based approach to MEMS sensors.