Piezoelectric transducers can serve as sensors or actuators. This study aims to employ Bayesian theory for monitoring the health state of piezoelectric-bonded plate structure. Due to the characteristic of easy excitation, the excited mode shapes are regarded as damage indicators. Firstly, two-dimensional finite element method was used to simulate the interested undamaged and damaged states of structures. The simulation data are regarded as the known sample results. Then combing these data with measured data, the confidence probability of each damage condition can be calculated by Bayesian formula. In this paper, an aluminum plate clamped on one side is taken as an example, and self-coded program is used for the numerical simulation. The identification accuracy of this method under the conditions of zero noise, Gaussian noise and impact noise is considered.