APLICATION OF FEEDFORWARD AND RECURRENT NEURAL NETWORKS FOR FLAPPING AND TORSION IDENTIFICATION OF A HELICOPTER BLADE

The identification of non linear systems is an experimental assessment, consisting of the development of techniques for dynamic systems models estimation from experimental data, demanding no previous knowledge of the procces. In this work a comparative study of the bilinear dynamics identification of a helicopter rotating single-blade mathematical model, carried through artificial neural networks, will be presented. The results of the identification are obtained by feedforward multilayer time delay neural networks (TDNN) and by recurrent neural networks. The feedforward multilayer networks that use delayed inputs are applied to problems involving secular processing, since the delayed signals form a memory device. The recurrent networks are networks that possess one or more feedback connections, being local if it is done at a neuron level or global if it encloses one or more complete layers. To illustrate the performance of the networks, the model, first will be executed conventionally and later with the neural networks. To compare the performance of the networks, the signal-noise relations of the estimated values by the networks will be calculated. In order to do a qualitative analysis, the return maps with the simulation results generated by the network will be plotted.

[1]  Dean W. Sparks,et al.  Design of Neural Networks for Fast Convergence and Accuracy , 1998 .

[2]  Gail D. Baura,et al.  Nonlinear System Identification , 2002 .

[3]  O. Nelles.  Nonlinear System Identification , 2001 .

[4]  Georgios B. Giannakis,et al.  A bibliography on nonlinear system identification , 2001, Signal Process..

[5]  L. A. Aguirre,et al.  IDENTIFICAÇÃO DE SISTEMAS NÃO-LINEARES UTILIZANDO MODELOS NARMAX POLINOMIAIS — UMA REVISÃO E NOVOS RESULTADOS , 1998 .

[6]  Garrison W. Greenwood,et al.  Training multiple-layer perceptrons to recognize attractors , 1997, IEEE Trans. Evol. Comput..

[7]  N. Saravanan,et al.  Modeling of the Space Shuttle Main Engine Using Feed-forward Neural Networks , 1993, 1993 American Control Conference.

[8]  Eric Hervieu,et al.  Identificação de regimes de escoamento bifasico horizontal com auxílio de redes neurais , 1999 .

[9]  Luciane de Fátima Rodrigues de Souza.  Identificação da dinâmica não linear de uma pá de helicóptero via redes neurais , 2002 .

[10]  Eduardo Morgado Belo,et al.  Identificação do comportamento não linear de um aerofólio flexível , 2001 .

[11]  I. J. Leontaritis,et al.  Input-output parametric models for non-linear systems Part II: stochastic non-linear systems , 1985 .

[12]  P. Werbos,et al.  Long-term predictions of chemical processes using recurrent neural networks: a parallel training approach , 1992 .

[13]  Hava T. Siegelmann,et al.  Computational capabilities of recurrent NARX neural networks , 1997, IEEE Trans. Syst. Man Cybern. Part B.

[14]  Dean W. Sparks,et al.  Neural Networks for Rapid Design and Analysis , 1998 .

[15]  I. Takahashi,et al.  IDENTIFICATION FOR CRITICAL FLUTTER LOAD AND BOUNDARY CONDITIONS OF A BEAM USING NEURAL NETWORKS , 1999 .

[16]  Peter Tiño,et al.  Learning long-term dependencies in NARX recurrent neural networks , 1996, IEEE Trans. Neural Networks.

[17]  Ah Chung Tsoi,et al.  Recurrent Neural Network Architectures: An Overview , 1997, Summer School on Neural Networks.

[18]  Roberto Celi,et al.  Recent Applications of Design Optimization to Rotorcraft-A Survey , 1999 .

[19]  Johan A. K. Suykens,et al.  Nonlinear system identification using neural networks , 1996 .

[20]  Richard D. Braatz,et al.  On the "Identification and control of dynamical systems using neural networks" , 1997, IEEE Trans. Neural Networks.