1. an inverse R prediction technique,
2. a polynomial function prediction technique,
3. and a nondimensional parameter prediction technique

The inverse R prediction technique uses a very general form of the prediction equation that can be applied in the same manner to all problems. Thus, the user is not required to develop a suitable form for the prediction and additional independent variables can be easily incorporated as needed. This method is designed to use a database that can be continuously updated as new experimental data becomes available. The method automatically takes advantage of the most appropriate data in the database for a given set of independent variables. The measured data points that are used for prediction are "radiating" information to the interpolation point. The farther the data point is away, the weaker the "radiation".

The polynomial function prediction technique is based on the concepts associated with the finite element method (FEM). In FEM, relatively low order polynomials are used to interpolate the functions of interest over a small portion of domain where the function is active called an element. The coefficients of the polynomial are derived from known values of the function of interest at points called nodes on the boundary of the element. For this application, the nodal values of the functions of interest were measured experimentally and are thus known quantities. This technique involves selecting a sufficient number of experimental data (node) points and then determining the coefficients of the polynomial from this data.

The nondimensional parameter prediction technique has been found to be the best method to represent empirical functions in many applications. The function coefficients are determined using an optimization routine to adjust the values of the coefficients so as to maximize the coefficient of determination of each of the functions. The nondimensional functions are adjusted to match the experimental results as closely as possible in a least squares sense. This approach to coefficient evaluation is suitable for any form of prediction function - linear or nonlinear.