An Adaptive Lattice Filter Approach to Structural, Failure Detection and Localization Based on Acoustic Reflections

A new approach to structural failure detection and localization is introduced based on acoustic reflections. It is shown that the cross-sectional areas of structural elements can be computed directly in terms of the reflection coefficients of an optimal finite impulse response Wiener filter realized in lattice form. This leads to an elegant method to detect and localize structured failures using recursive on-line estimation methods. There are many advantages of this approach relative to standard failure detection schemes. Specifically, the acoustic reflection approach: 1) does not require training; 2) does not require prior knowledge of a structural model; 3) can detect and localize multiple failures; 4) can indicate the extent of damage at each location. A simulation example is given which successfully demonstrates each of these qualities. For the present paper, the acoustic reflection method is established by working out the theory completely for a bar with nonuniform cross-sectional area in axial vibration. Some ideas for extending the theory to more elaborate and realistic structural configurations are briefly outlined.