Анализ Гибридных Регуляторов В Моделях Управления Техническими Объектами В Изменяющихся Условиях

Abstract

paper analyzes hybrid controllers for control models of technical objects operating in changing conditions. It also considers the models which involve control based on hybrid controllers implemented on the basis of sequential interaction between PI-and IPI-FUZZY-controllers and PID-and IPD-FUZZY-controllers with the generated structure of the Sugeno-type fuzzy inference system and the developed ANFIS model.In hybrid controllers, the fuzzy controller rule base is formed automatically using a specially developed algorithm based on data obtained from a classical controller with subsequent training using a neural network. The ANFIS design principle in the form of a hybrid network for PI and IPI-FUZZY controllers is the use of the output signal error indicators, its integral (differential for PID and IPD-FUZZY controllers) and control action. The following aspects have become the development features. In order to test the hybrid network efficiency to identify the fact of its retraining, the authors used the data obtained as a result of the classical regulator operation; to form a training sample for building a hybrid network they used the data obtained as a result of the fuzzy regulator operation. This makes it possible to exclude expert's participation in the synthesis of the fuzzy controller rule base and to ensure efficient and robust control of an object functioning in unforeseen external situations.The IPI-FUZZY-controller and the IPD-FUZZY-controller have shown better quality indicators comparing to the corresponding classical ones, which makes it possible to recommend using in real control systems. The presented models were developed in the Simulink environment and the ANFIS editor of the Fuzzy Logic Toolbox extension package.