To Identify a Torque Controller System Approximating a Neural Network Based on Model Reference Technique

Abstract

In this paper, a system of a torque controller assembly, using a split field winding dc motor, controls the sliding movement of an object, is proposed. This system is identified using the neural network (NN) based model reference technique. At first, a neural network, called plant identification neural network controller (PINNC) is configured by learning the behavior of the torque controller assembly through training the network. The 'process of learning' uses a method of training (trainlm) to train the network. Then the outputs of the proposed system and this PINNC are compared. The knowledge of this comparison is given as feedback to a second neural network controller, called NN-plant or neural network model reference controller (NNMRC). The NNMRC is configured by learning the behavior of a reference model system, provided to it. The 'process of learning' uses a method of training (trainbfgc) to train this network. The plant identification neural network controller (PINNC) drives the neural network model reference controller (NNMRC) and controls the output of the NNMRC. Ultimately, the neural network model reference controller (NNMRC) or NN-plant identifies the proposed system, the torque controller assembly reliably and successfully.