A parametric investigation of state-space-based prediction error methods with stochastic excitation for structural health monitoring

The dynamic response of structures has often been used as a basis for analysis within the field of structural health monitoring. Recent work has presented the use of the state-space representation of the structural response to extract damage sensitive features. One particular feature involves using a baseline attractor to predict the evolution of an attractor at some later time, and using the prediction error to determine the presence and extent of damage. Such state-space-based methods are reliant on a number of parameters related to excitation type, reconstruction of the attractor, and prediction technique. This work investigates the effect of the various parameters on prediction error results and the ability to detect damage for both a computational model and an experimental test structure, with emphasis placed on the use of a band-limited stochastic excitation.