Principal Component Regression for Fitting Wing Weight Data of Subsonic Transports

This paper documents the lessons learned from fitting the wing weight data of 41 subsonic transports by a semi-empirical regression model, least polynomial interpolation, radial basis function interpolation, Kriging interpolation, Gaussian process, and principal component regression using radial basis function interpolation. The paper discusses various aspects of fitting data to a wing weight model: data scaling, variable selection, principal component analysis, subjective choice of input variables, interpolation methods, and verification of constructed wing weight models. The numerical results show that principal component regression using multiquadric radial basis function interpolation is capable of capturing physical trends buried In the wing weight data and generates the most useful wing weight model for conceptual design of subsonic transports among the tested data fitting methods. Even though the benefits of principal component regression are only demonstrated by the wing weight data fitting problem, the methodology could have significant advantages in fitting other historical or sparse data.