Identification of aerodynamic coefficients of ground vehicles using neural network

The purpose of this paper is to demonstrate the application of a combination of neural network and an oscillating model facility as an approach in identification of aerodynamic coefficients of ground vehicle. In literature study, a method for estimating transient aerodynamic data has been introduced and the aerodynamic coefficients are extracted from the measured time response by means of conventional approach. The potential of neural network as an alternative method is explored. For simplicity, only the damped oscillation considered in this analysis while neglecting any unsteadiness or buffeting load. Two feedforward neural networks are constructed to estimate the damping ratio and natural frequency, respectively, from the measured time response recorded during the dynamic wind tunnel test. These two parameters are used to calculate the aerodynamic coefficients of the ground vehicle model. To validate the network approach, the resulted coefficients are compared with the ones retrieved conventionally. By simulating the system's transfer function, the response generated from neural network results were found to be closer to the measured time response compared to the response generated using the conventionally estimated coefficients.