Flow variance method for damage identification
Date of Original Version
A new method called flow variance for multidimensional damage identification is introduced. It can be considered a simplified practical implementation of the phase space warping concept. In a system, subjected to damage accumulation, parameters drift causing the alteration of the trajectory flow in its phase space. A flow variance metric is used to quantify these changes in the velocity of the flow. In an experimental context, a phase space trajectory is reconstructed based on delay coordinate embedding of measured scalar time series. In this reconstructed phase space the flow variance metric is calculated by estimating the shift in the trajectory's average evolution time rate of change in small disjoint sections of the reconstructed phase space. It is hypothesized that a feature vector composed of flow variance metrics can be projected to actual damage states. An experimental system of a harmonically driven cantilever beam in a double-well potential force field is used to validate this method. Two-dimensional damage process is introduced through electromagnets altering the beam's potential field. Demonstrating the existence of the affine projection between the estimated and independently measured damage states validates the hypothesis. The smooth orthogonal decomposition based damage identification is also able to identify and recover the active damage states. This method is simple to implement and requires modest computational resources.
Conference Proceedings of the Society for Experimental Mechanics Series
Liu, Ming, and David Chelidze. "Flow variance method for damage identification." Conference Proceedings of the Society for Experimental Mechanics Series , (2005). https://digitalcommons.uri.edu/mcise_facpubs/114