“Cut-Glue” Approximation in Problems on Static and Dynamic Mathematical Model Development

Abstract

The solution to the problem on building mathematical models of technical objects through the approximation of various experimental dependences is offered in the paper. This approach is especially true for modeling aircraft because the aerodynamic coefficients of their models can be obtained either by full-scale study or by computer simulation only. Currently, the experimental simulation is performed either through the regression analysis (RGA) methods, or through spline approximation. However, the RGA has a significant disadvantage, namely a poor approximability of piecewise and multiextremal dependencies. The RGA gives a rough approximation of the experimental data for similar curves. Spline approximation is free from this disadvantage. However, a high degree of discretization, a strict binding to the number of spline points, and a large number of equations, make this approach inconvenient for application when a compact model building and an analytic transformation are required. A problem solution combining the advantages of both approaches and clearing up the troubles is offered in the paper. The proposed approach is based on the regression construction of the mathematical models of the dependence fragments, the multiplicative excision of these fragments in the local functional form, and on the additive combining of these local functions into a single analytic expression. The effect is achieved by using special “selection” functions multiplicatively limiting a nonzero definition domain for each of the approximating functions. The method is named “cut-glue” by the physical analogy of the approximation techniques. The order and structure of the approximating function for each segment can be arbitrary. A significant advantage of the “cut-glue” approximation is in a single analytic expression of the whole piecewise function instead of a cumbersome system of equations. The analytical and numerical studies of the properties and operational experience of the proposed method are resulted.