A subspace-based approach to structural health monitoring

The authors introduce a subspace-based method for tracking structural changes in mechanical systems. Most existing signal processing methodologies in the realm of modal analysis based on vibration signals acquired from accelerometers fail to move beyond merely detecting the existence of changes. Such changes are revealed through departures of modal parameter estimates (eigenfrequencies and mode shape vectors) from baseline values. Much less effort has devoted to diagnosing shifts in modal parameter estimates in terms of changes in mass, stiffness, or damping. In this paper, an advanced approach for detecting and diagnosing (i.e., isolating) changes in aircraft and other mechanical structures is presented. The approach builds upon a broad base of established theoretical methods, including modern spectral analysis (autoregressive, moving-average modeling) techniques for modal parameter estimation and presents a new theoretical approach for diagnostic tracking. The algorithmic methodology is illustrated via simulation of a simple mechanical system. The methods presented lend themselves to nonintrusive, in situ monitoring based on natural excitation and are retrofittable to a broad class of existing structures in most immediate need of effective structural health monitoring tools.