A simulation-based inverse design of preset aircraft cabin environment

The inverse design of preset aircraft cabin environment (ACE) is presented. Five design variables (inlet velocity, angle, temperature, position and outlet position) and three design objectives (PMV, DR and Air Age) are involved in the current inverse design. The Artificial Neural Network (ANN) and genetic algorithm (GA) are combined to design ACE based on the Computational Fluid Dynamics (CFD) analysis. To eliminate the uncertainty and risk of accumulative errors in the design process, both ANN and CFD are used to obtain the design objectives of new individuals generated by GA. To enhance the prediction accuracy of ANN, three single-output ANNs for each design objective are adopted instead of one mutiple-output ANN. The results obtained by GA alone and the proposed method are compared. Instead of applying GA, 57% of computational costs are reduced when the proposed method is used. Comparing the design results of different scales of CFD databases, it is found that the CFD database with 110 samples has the less computational cost, while that with 70 samples has better solutions.