본 논문에서는 KC-100 비행시험데이터를 인공신경망에 학습시킨 뒤, Partial Differential Method(PDM)을 적용하여 종축 공력미계수를 추정하였다. 고전적인 파라미터 추정기법인 최대공산 추정법 MMLE와 선형⬝ 비선형 반응 비교를 통해 PDM의 유효성을 검증하고, 기존에 사용되던 고전적인 파라미터 추정기법의 대안으로 제시하였다.
In this paper, the aerodynamic stability and control derivatives of the KC-100 airplane are estimated using the Partial Differential Method(PDM) based on the artificial neural network. To verify the estimation accuracy of PDM, it is compared with Modified Maximum Likelihood Estimation(MMLE), a classical parameter estimation method. MMLE uses the aerodynamic DB(database) as the initial estimation value, and the artificial neural network does not need the initial value for estimation. Therefore, the estimation accuracy of PDM can be checked by comparison with the MMLE method which is believed to be more accurate. The artificial neural network has a FeedForward Neural network structure, in which two hidden layers are fixed but the number of neurons can be selected to optimize the learning efficiency. The hyperbolic tangent sigmoid function is used as the activation function of the hidden layer. Once the artificial neural network structure is optimized, the aerodynamic stability and control derivatives are estimated using PDM. The aerodynamic DB is tuned with the estimated value of PDM. It is used to check aircraft's linear and nonlinear responses, which are also compared with those based on the aerodynamic DB tuned with the MMLE method. This comparison confirms that the proposed PDM is effective and can be an alternative to the classical parameter estimation method.