An Aggregate Multi-loop Control for a Dc Servodrive. the Design of the Neuro-fuzzy Controller

The purpose of the study is the synthesis of a reliable and robust fuzzy controller for a 2 order DC servodrive. A first step is to prepare the computations associated to the main control loops for the above mentioned servosystem. This approach aims to aggregate the relations for the position and the velocity loops as an unique computation formula expressed into the variables specific for the real-time processing, optimized and adapted to a microcontroller arithmetic. The systemic variables considered are the position error and its variation during a sampling period, this second input into the next neuro-fuzzy control having the physical meaning of the speed. Both variables are expressed as dependences of the number of encoder impulses delivered during the sampling time. A second step is the design of a neuro-fuzzy controller for both the control loops. The ANFIS method uses training data delivered by the well tuned conventional control structure for generating a DISO fuzzy controller. Some special blocks were added to the standard structure in order to improve the IN-OUT data quality. Appreciations are made for the quality of the training process and the characteristics of the fuzzy inference system are presented. A subsequent paper will analyze the system with the neuro-fuzzy controller and its real time-time implementation.