Determining the Service Life of a Metal Shell Structure Using a Neural Network

An effective method for determining the service life of a metal shell structure under variable thermomechanical load is proposed. The stress-strain state of the structure is determined by solving a nonlinear boundary value problem of thermoplasticity for a thin-walled shell of revolution. The service life of structures under variable thermomechanical loads is determined based on the equation of low-cycle fatigue of the material. The application of the presented method is demonstrated using the example of a shell structure designed for high-temperature annealing of electrolytic steel. The temperature of the protective shell during operation can reach 1000 °C. However, this shell is made not of heat-resistant metal, but of St3 steel, which is not intended for use at such an extreme temperature. In the absence of the necessary mechanical parameters for the material of the structure at high temperatures, the prediction method is used. The found service life values for the metal shell with the mechanical parameters of the material at known temperatures are extrapolated to higher operating temperatures using neural networks.