In many engineering problems involving optimization of a function or fitting experimental data to a function of several variables, analytical procedures fail, either due to the complexity of the problem or difficulties involved in obtaining analytical form for partial derivatives. When the number of independent variables of the function to be optimized gets large, and if the function has many nonlinearities and local optima, the well-known optimization methods1#x2032;3#x2032;4 fail to guarantee a global optimum for the function. This is often encountered in functional-fit problems where one tries to minimize the least square of the error between experimental measurements and values of the function. In such cases one invariably has to resort to a search procedure.In this note we describe a search procedure to be used on a digital computer for either fitting experimental observations to a nonlinear function of several variables or optimizing a function having several independent parameters.
展开▼