Deep Neural Network Model-based Parameter Estimation Method Using Transmissibility of Cantilevered Beam

In this study, we developed a method for estimating mechanical properties of cantilever beams, such as elastic modulus and damping coefficient, using a deep neural network model. Analytically, the transmissibility at the tip of the cantilever modeled using Euler-Bernoulli beams was used as training data for the deep neural network model. In addition, the robustness of the proposed method was examined by adding Gaussian noise to the transmissibility to investigate the effect of noise that may have occurred in the experiment. We demonstrated that the deep neural network model estimates unknown parameters with high accuracy, even in the presence of noise. Finally, the estimation results of unknown parameters for various initial values were compared using the proposed method, gradient descent method, and pattern search method.