Viscoelastic analysis of a sleeve based on the BP neural network

Viscoelastic materials are widely used in aviation, underground engineering, chemical engineering, light textile, construction, machinery and shock-absorbing. The properties of viscoelastic materials present strong non-linear characteristics in the time, temperature and frequency domains. The objective is to explore the possibility of having a rational understanding of the complex response. The neural network technique is thus implemented for this purpose. This study presents the differential constitutive equations of viscoelastic materials based on the generalized Kelvin model. The viscoelastic response Laplace transform equations of the sleeve structure are derived under the condition of plane strain according to the boundary conditions of the viscoelastic sleeve. The orthogonal design test results of viscoelastic numerical calculations are normalized and trained using a three-layer BP neural network. The trained network can easily realize the mapping of the viscoelastic material test set. Simulation results show that the neural network test and numerical solution results are close to each other, and the maximum error of all variables is less than 2.8 percent. The results obtained in this research reveal that the neural network technique has a suitable capability in establishing correlations among different loads and corresponding viscoelastic responses. The developed neural network model can effectively solve the non-linear problems of viscoelastic structures.