Study of new applications of nonderivative analog optimizers

Abstract

In this project of scientific initiation will be approached the study of new applications of non-derivative analog optimizers. For this, a study will be carried out on the non-derivative analog optimizer proposed by Teixeira and Zak (1998), analyzing its operation in the search for the optimum point. The optimizer Teixeira and Zak can be classified as an artificial neural system and is said to be one-dimensional because it seeks the minimum of a function with only one variable. It will be done a survey of the new applications of non-derivative analog optimizers, proposing the mathematical modeling of these applications, so that later the simulation of them is done. As an example, known applications of these non-derivative optimizers are: at the field of industrial process control; navigation of underwater vehicles; in advanced control system design for Antilock Brake System (ABS); in the control of the Power Factor (PF); at the search of the Maximum Power Point Tracking (MPPT) and in the control of uncertain systems using a switched controller for plants with uncertain control direction. In the mentioned applications, we try to optimize (minimize or maximize) an objective function, with only a minimum (or maximum) local or global, without knowing this function perfectly. In this situation, non-derivative analog optimizers are very useful because they do not use the objective function gradient. It is known that the optimizer Teixeira and Zak is not restricted only in the search for the optimal point, it can be adapted according to the need of the objective function in order to achieve satisfactory results. Thus, for the validation of the project will be carried out simulations of the applications proposed in the software Matlab and Simulink, evaluating the response of the optimizer and modeling it according to the dynamic system in question. (AU)