Electromechanical impedance based self-diagnosis of piezoelectric smart structure using principal component analysis and LibSVM

Abstract The long-term use of a piezoelectric smart structure make it difficult to judge whether the or lead zirconate titanate (PZT) is damaged when signal changes. If sensor fault occurs, cases and degrees are unknown based on electromechanical impedance method. Therefore, after principal component analysis (PCA) six characteristic indexes, two-component solution that could explain 99.2% variance in original indexes was obtained damage comes from PZT. Then LibSVM used an effective identification four faults (pseudo soldering, debonding, wear, breakage) their three degrees. result shows accuracy PZT reached 97.5%. absolute scores PCA comprehensive evaluation for structural damages less than 0.5 while greater 0.6. By comparing samples under conditions with set threshold, occur effectively judged; intact 12 possible states can be all classified correctly model trained by LibSVM. It feasible classify faults.