Damage Identification Method Based on PZT Admittances and a Multi-Task Network Model

A damage identification method based on a multi-task deep learning network and Bayesian fusion is proposed in this paper. Firstly, the 3D finite element model of a cantilever steel beam with three surface-bonded piezoelectric ceramic (PZT) transducers is established and validated experimentally. The admittance signals of PZT under different damage scenarios will be obtained by simulating the damage with a cut on the beam. Next, the admittance data are input into the constructed multi-task network model for training and validation. Then, the test data are input into the trained model, and the damage of a steel beam is located and quantified to verify the effectiveness of the proposed method. Finally, Bayesian fusion is used to fuse the identified results based on single PZT admittance signals at different positions to improve the accuracy of damage identification.