In this paper, a flow variance based structural damage identification method is introduced. It is a practical implementation of <i>phase space warping</i> concept as applied to damage identification. A coupled dynamical system is considered where a slow-time damage process causes drifts in the parameters of fast-time system describing measurable response of a structure. The method is based on the hypothesis that the distribution function of the fast-time trajectory is a function of damage state. In this method, estimated local expectation of trajectory in its phase space is used as a damage tracking feature vector. After the feature vectors are constructed, damage identification is realized by <i>smooth orthogonal decomposition</i>. Data processing time requirements of the flow variance based approach decrease by about 2 orders-of-magnitude compared with the phase space warping based method. A common form of the feature vectors is also discussed.