Fast Prediction of Turbine Rotor Temperature Field based on Deep Neural Network and Fatigue Life Early Warning

Nowadays, high-capacity thermal power plants need to strengthen the ability of fast start-stop and peak load regulation, but frequent start-stop and load changes cause the rotor to withstand alternating thermal stress, which is easy to produce low-cycle fatigue damage, and seriously affect the safety of the unit. Therefore, it is very important to predict the temperature field of turbine rotor in real time and evaluate its overall life to ensure the safe operation of turbine. In this paper, the simulation data of the Computational Fluid Dynamics (CFD) is firstly preprocessed. Then the prediction model of turbine temperature field based on deep neural network is constructed. Finally, the predicted temperature field is converted into thermal stress, and the thermal stress of each point is brought into the rain flow method to calculate the amplitude, average value and cycle number, and then the results are input into the fatigue curve formula to establish a "thermal stress-cycle life" model, so as to give early warning to the overall life of the rotor.