A pattern recognition approach for identification of transducer-structure debonding using Lamb waves

In structural health monitoring, using piezoelectric transducers to generate high frequency elastic waves like Lamb waves in the structures is eminent. In general, piezoelectric transducers are assumed to be perfectly bonded with the host structure; however, in practical environment, there are possibilities for them to have faults. Since detecting, locating and assessing damages in a structure depend solely on transducer responses, transducer fault identification is vital. By using electrical admittance, axial strain and shear stress as function of frequencies or in analog interface circuits to identify faulty transducers, lead to demand of circuitry and processes, consequently increasing the implementation complexity. Hence, we propose a pattern recognition system that can identify transducers that are partially bonded to host structure. This pattern recognition system employs classification for features extracted from instantaneous Lamb wave signals with no need of baseline data.