Damage Assessment using Generalized State-Space Correlation Features

Recently, damage detection capability has been demonstrated successfully using state-space based algorithms. These methods are advantageous because they rely on data-driven techniques that do not conform to models or assumptions like linearity. State-space-based features traditionally involve comparisons between measurements taken at the same location but at different times to determine if a change has taken place. However, if features such as state-space cross-prediction error and generalized interdependence are formulated such that they instead employ comparisons between simultaneous measurements at different locations, a fuller assessment of structural damage is possible. In addition to the presence of damage, other characteristics such as the extent, location, and type of damage can be revealed from these features. This approach is validated through a multi-degree-of-freedom oscillator and an experimental frame structure.