Support vector machine approach to drag coefficient estimation

An effective artillery attack depends on the accuracy of the firing tables of guns, and firing table is essentially related to the drag coefficient, which determines the ballistic behavior of the projectiles. In this paper, the estimation of drag coefficient is treated as a regression problem in the machine learning setting and application of support vector regression is proposed as the solution. Separate Support Vector Machines (SVMs) are used for subsonic, transonic and supersonic regions. The inputs to the SVMs consist of physical parameters and Mach number and the output is the zero-yaw drag coefficient. The use of SVM with sufficient training data set obviates the need for post-estimation corrections with real firing data. The method of Support Vector regression can be used for the estimation of other aerodynamic coefficients also.