Training wavelet networks for nonlinear dynamic input–output modeling

Oussar, Yacine; Rivals, Isabelle; Personnaz, L.; Dreyfus, Gérard

doi:10.1016/s0925-2312(98)00010-1

Cited by 182 publications

(94 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A four layers WNN is introduced as shown in Fig. 1, which is comprised of an input layer, a mother wavelet layer, a wavelet layer, and an output layer [17][18][19]. The basic principle and function of each layer are specified in the following section.…”

Section: Adaptive Wnn Uncertainty Observermentioning

confidence: 99%

Adaptive Wavelet Neural Network Backstepping Sliding Mode Tracking Control for PMSM Drive System

Liu

2014

Automatika

View full text Add to dashboard Cite

Original scientific paperThis paper presents a wavelet neural network backstepping sliding mode controller (WNNBSSM) for permanentmagnet synchronous motor (PMSM) position servo control system. Backstepping sliding mode (BSSM) is utilized to guarantee favorable tracking performance and stability of the whole system, meanwhile, wavelet neural network (WNN) is used for approximating nonlinear uncertainties. The designed controller combined the merits of the backstepping sliding mode control with robust characteristics and the WNN owning the capability of artificial neural networks for online learning and the capability of wavelet decomposition for identification. An observed error compensator is developed to compensate the estimated error of the WNN and the adaptive law is derived according to Lyapunov theorem. The effectiveness of the proposed controller is investigated in simulation under different operating conditions. The simulation results demonstrate the proposed WNNBSSM controller can provide precise tracking performance and robust characteristics despite unknown parameter uncertainties and load disturbance. Moreover, an implemental wavemaker system is established to verify the effectiveness of the proposed control algorithm.Key words: Backstepping, PMSM position servo system, Robustness, Sliding mode control, Wavelet neural networkAdaptivno slijedno upravljanje u kliznom režimu rada s valićnom neuronskom mrežom za sustav s PMSM elektromotornim pogonom. U radu je predstavljen povratnokoračni regulator u kliznom režimu rada za servo sustav pozicioniranja s motorom s permanentnim magnetima (PMSM) zasnovan na valićnim neuronskim mrežama (WNNBSSM). Povratnokoračni klizni režim rada (BSSM) korišten je za jamčenje željenih performansi slije enja i stabilnost cijelog sustava, dok je valićna neuronska mreža (WNN) korištena za aproksimaciju nelinearnih nesigurnosti. Sintetizirani regulator povezuje prednosti i robusne karakteristike povratnokoračnog kliznog režima upravljanja te WNN sa sposobnostima umjetnih neuronskih mreža za online učenje i mogućnost valićne dekompozicije za identifikaciju. Razvijen je kompenzator pogreške estimacije WNN i adaptivni upravljački zakon prema Ljapunovljevom teoremu. Djelotvornost predloženog regulatora istražena je kroz simulacije u različitim uvjetima rada. Rezultati simulacija pokazuju da predloženi WNNBSSM može osigurati precizna svojstva slije enja i robusne karakteristike unatoč nepoznatim nesigurnostima parametara i poremećaja tereta. Uz to, razvijen je izvedbeni sustav generatora valova za provjeru djelotvornosti predloženog upravljačkog algoritma.Ključne riječi: povratnokoračni regulator, PMSM servo sustav za pozicioniranje, robusnost, upravljanje s kliznim režimom rada, valićna neuronska mreža

show abstract

Section: Adaptive Wnn Uncertainty Observermentioning

confidence: 99%

Adaptive Wavelet Neural Network Backstepping Sliding Mode Tracking Control for PMSM Drive System

Liu

2014

Automatika

View full text Add to dashboard Cite

show abstract

“…In order to approximate arbitrary nonlinear functions the wavelet network which combines feedforward neural networks and wavelet decompositions has been proposed in [9]. The identification of static and dynamical systems using wavelet network have attracted many researchers since the wavelet analysis has been successfully applied for analyzing signals both in time and frequency domain with different resolution levels [10][11][12][13].…”

Section: System Modellingmentioning

confidence: 99%

“…Therefore, the next state of the system is predicted from the previous observations. The system evolution function F is approximated by some arbitrary set of basis functions for the modelling or identification of the nonlinear dynamical systems [10][11][12]. The purpose is to represent Eq.…”

Section: Dynamical System Modelling With Wavelet Networkmentioning

confidence: 99%

“…Also, there have been studies on the reconstruction of the attractors of the chaotic systems using the Wigner-Ville distribution [7] and the wavelet transform [8]. The wavelet network [9] which combines feedforward neural networks and wavelet decompositions, have attracted many researchers since the wavelet analysis has been successfully applied for analyzing signals both in space and frequency domain with different resolution levels [10][11][12]. Additionally, the wavelet network is powered with the statistical techniques in [13].…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

The system modelling and circuit implementation from time–frequency domain signal specifications

Savacı

Özkurt

2008

AEU - International Journal of Electronics and Communications

View full text Add to dashboard Cite

“…Wavelet networks have been used both for static [10,12] modeling and for dynamic input-output modeling [9]. It was proved in [4] that families of wavelet functions -particularly wavelet frames -are universal approximators, which gives a theoretical basis to their use in the framework of function approximation and process modeling.…”

Section: Introductionmentioning

confidence: 99%

Initialization by selection for wavelet network training

Oussar¹,

Dreyfus²

2000

Neurocomputing

170

108

View full text Add to dashboard Cite

We present an original initialization procedure for the parameters of feedforward wavelet networks, prior to training by gradient-based techniques. It takes advantage of wavelet frames stemming from the discrete wavelet transform, and uses a selection method to determine a set of best wavelets whose centers and dilation parameters are used as initial values for subsequent training. Results obtained for the modeling of two simulated processes are compared to those obtained with a heuristic initialization procedure, and demonstrate the effectiveness of the proposed method.

show abstract

Training wavelet networks for nonlinear dynamic input–output modeling

Cited by 182 publications

References 11 publications

Adaptive Wavelet Neural Network Backstepping Sliding Mode Tracking Control for PMSM Drive System

Adaptive Wavelet Neural Network Backstepping Sliding Mode Tracking Control for PMSM Drive System

The system modelling and circuit implementation from time–frequency domain signal specifications

Initialization by selection for wavelet network training

Contact Info

Product

Resources

About