AN IMPROVED NUMERICAL-METHOD TO FIND PARAMETERS USING EXPERIMENTAL-DATA

In this paper we describe a numerical method to determine adjustable p arameters in a mathematical model using experimental data. This is bas ically an improvement over our earlier method which was based on the i dea that the area under the experimental curve must be equal to the in tegral of the function used. The improvement used is the addition of a n area correction factor which estimates the necessary difference that must exist between the numerically evaluated and the true area. This correction surprisingly eliminates the use of integral with the result that the two areas being equalized are both numerically evaluated, on e using the experimental data points and the other using the fitted fu nction values. It is shown that the application of the area correction factor significantly improves the accuracy of the adjusted parameters . The method has been compared with the well-known method of least squ ares for few selected cases involving variety of functions. It is seen that our method shows convergence for a wider range of initial guesse s as compared to the method of least squares. The superiority of our m ethod becomes evident when more than two non-linear parameters are inv olved.