Assistant investigation of energy dissipation in non-obstructive particle damper based on a neural network using simulated annealing backpropagation

Abstract

The particle damper has been widely used as an efficient passive vibration control device in recent years. The highly non-linear characteristic of the energy dissipation mechanism is essential for our increased understanding of non-obstructive particle damper (NOPD). To connect motion modes of the granular system and energy dissipation, we developed a neural network using simulated annealing backpropagation (SA-BP) to predict the loss factor of NOPD in this paper. The simulations based on the verified discrete element method (DEM) model are carried out, and the data is used to train and test the neural network. Based on the prediction of a well-trained neural network using SA-BP, the relationship between the loss factors and the rheology behaviors of the granular system is discussed. This paper effectively combines intelligent algorithms (SA-BP) and particle damping characteristics. The algorithm is expected to be further used as an auxiliary method for the experimental study of the granular system.