Artificial neural network-based design tool for the horizontal stabilizer of a helicopter

Designing a helicopter is quite difficult and complex. Every decision made in the preliminary design phase seriously impacts the further design phases of a helicopter project. Therefore, it is vital to initially make the right and logical decisions for the design. Hand calculations, finite element analyses, and structural tests are beneficial to determine the conceptual design parameters. However, it takes much time to perform many finite element analyses and hand calculations. Furthermore, testing different types of structures in the early design stages can be very costly. These time-consuming and expensive processes are repeated in the structural design stage of each various project. Therefore, an efficient solution is needed to address this problem. Artificial neural networks are powerful models that may be used to overcome this issue and to reduce the effort and time spent in the initial design phase. In this thesis, an artificial neural network-based design tool has been developed to determine the static structural characteristics of the horizontal stabilizer of a helicopter. The database required for training an artificial neural network model was created utilizing the finite element analyses of the horizontal stabilizer. These analyses were performed under the aerodynamic load for different design variables. The neural network model trained with this data was built in Python using the Keras library. The model outputs were then compared with the finite element analysis results, and the model performance was presented. Lastly, the database was reduced using the Hammersley sampling methodology, and the effect of decreasing the number of data feeding the network model was evaluated.