NONLINEAR SYSTEM MODELING USING SINGLE NEURON CASCADED NEURAL NETWORK FOR REAL-TIME APPLICATIONS

Abstract
Neural Networks (NN) have proved its efficacy for nonlinear system modeling. NN based controllers and estimators for nonlinear systems provide promising alternatives to the conventional counterpart. However, NN models have to meet the stringent requirements on execution time for its effective use in real time applications. This requires the NN model to be structurally compact and computationally less complex. In this paper a parametric method of analysis is adopted to determine the compact and faster NN model among various neural network architectures. This work proves through analysis and examples that the Single Neuron Cascaded (SNC) architecture is distinct in providing compact and simpler models requiring lower execution time. The unique structural growth of SNC architecture enables automation in design. The SNC Network is shown to combine the advantages of both single and multilayer neural network architectures. Extensive analysis on selected architectures and their models for four benchmark nonlinear theoretical plants and a practical application are tested. A performance comparison of the NN models is presented to demonstrate the superiority of the single neuron cascaded architecture for online real time applications.

Authors
S. Himavathi1, A. Muthuramalingam2, A. Venkadesan3 and K. Sedhuraman4
Pondicherry Engineering College, India,

Keywords
Single Neuron Cascading, Neural Networks, Modeling, Compact Models, Real-time Applications
Published By :
ICTACT
Published In :
ICTACT Journal on Soft Computing
( Volume: 2 , Issue: 3 )
Date of Publication :
April 2012

Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.