A genetic algorithm solution to the maximum likelihood statistical problem

The paper describes an application of a genetic algorithm (GA) computer software called Gen S, developed at CEPEL, with the support of major Brazilian utilities, in the solution of a complex statistical problem involving the maximization of the so called likelihood function. The problem, as described below, is associated with the complete statistical characterization of a given dielectric insulation, such as an air gap, from just a few observations of the gap behavior under certain stress conditions (high voltage signal applied to the gap a limited number of times). Besides comparing the results with a modified Newton-Raphson method, the authors point out the ease of use of Gen S as well as its major advantages regarding analytical and more conventional numerical methods.