Abstract
In this paper, a revised GMDH (Group Method of Data Handling) algorithm, which can identify a nonlinear exponential type model, is developed. In previous GMDH algorithms, it is proposed to use the functions such as polynominal, Bayes function, periodic function and rational function as the complete descriptions of the system. But, for the systems whose output variables variate smoothly and whose characteristics can be described by using the combinations of the exponential function, it is necessary to construct the complete description of the system by using the exponential function. So these systems have not been able to be identified acculately by using the previous GMDH algorithm. The revised GMDH in this paper generates some nonlinear exponential type models, which are constructed by two intermediate variables in each selection layer, and the complete description of the system is constructed by combining these nonlinear exponential type models in multilayered structure. The revised GMDH algorithm is applied to the identification problem of a furnace heat pattern and the result obtained is compared with those obtained by the previous GMDH algorithms in which the high order nonlinear polynomial is used as a complete description of the system.
