Thermal management of turbine disc cavity system using FFBPNN and NSGA II algorithm

Abstract

This study addresses the thermal management of the turbine disc cavity system (TDCS) by combining the feed-forward backward propagation neural network (FFBPNN) with the non-dominated sorting genetic algorithm II (NSGA II). First, the heat transfer analysis of the TDCS is carried out using the cross-scale computational model which is consist of 1D fluid network method and the 2D finite element method. The impact of different cooling air inlet conditions on the heat transfer performance of the TDCS is investigated. Results show that changing the inlet pressure and temperature significantly affects the heat transfer performance of the TDCS, and the TDCS temperature field can be regulated by the inlet parameters. Then, the prediction model based on the FFBPNN is established to predict the heat transfer performance of the TDCS, and satisfactory result is obtained with mean relative error lower than 1.5 % and a coefficient of determination higher than 0.998. Finally, the NSGA II is employed to optimize the cool air inlet condition to achieve thermal management of the TDCS. The Pareto solution set and the optimal solution are obtained. The results indicate that the most comprehensive improvement in the heat transfer performance of the TDCS can be achieved by present technology.