Adaptive Wavelet CMAC Tracking Control for Induction Servomotor Drive System

Vol.:(0123456789) 1 3

Journal of Electrical Engineering & Technology

https://doi.org/10.1007/s42835-018-00029-1

ORIGINAL ARTICLE

Adaptive Wavelet CMAC Tracking Control for Induction Servomotor

Drive System

Thanh‑Quyen Ngo1

 · Mien‑Ka Duong1

 · D. C. Pham1

 · Duc‑Nam Nguyen2,3

Received: 22 September 2017 / Revised: 2 January 2018 / Accepted: 28 August 2018

© The Korean Institute of Electrical Engineers 2019

Abstract

This research proposes an enhanced control system for induction servomotor to obtain the high-precision position tracking

based on wavelet cerebellar model articulation controller. The proposed controller combines the wavelet cerebellar articula￾tion model (WCMAC) with compensator controller to have high performance for the system. The superior properties of the

WCMAC are its fast learning of the CMAC and wavelet decomposition capability. Therefore, it is used to mimic a model￾based controller with exactly unknown parameters. The compensator controller with an estimated boundary error attenuates

the efects of disturbances due to time-varying parameters and load acting on the shaft of motor. The online learning rules

of WCMAC derived from gradient-descent method and Lyapunov function is used to estimate bound error to ensure the

stability of the system. The experimental results for induction servomotor are provided to conclude the efectiveness of the

proposed control system even the dynamic model of the induction servomotor is completely unknown.

Keywords Wavelet · Cerebellar model articulation controller (CMAC) · Uncertain non-linear systems · Servomotor ·

Neural network (NN) · Wavelet neural network (WNN)

1 Introduction

The feld-oriented control methods have been used since

the last decade to control induction motor drivers and have

high-performance [1, 2]. In these researches, the dynamic

behaviors of the induction motor are the same as a sepa￾rately excited DC motor. The rotor fux is assumed to be con￾stant and there was not mechanics coupling. Consequently,

the uncertainties of the plant, such as varying-mechanical

parameters, disturbances of external load in the practical

applications are not completely considered in the stable

problems. To overcome these problems, many intelligent

methods have been applied to control the induction servo￾motor driver systems [3–5]. Firstly, a model-based fuzzy

adaptive method is proposed by Liaw and Lin to attenuate

the efects of varying parameters [3]. However, the fuzzy set

rules must be constructed by a trial-and-error turning proce￾dure and this wastes lots of time. Secondly, the sliding-mode

control for the rotor fux torque is proposed by Chan and

Wang [4] to deal with disturbances and uncertainties of the

plant but the model-based algorithm is a limitation of this

method. In addition, novel techniques are proposed by Zig￾mund and Shieh to estimate rotor time-constant for indirect

feld-oriented induction motor drive [6, 7]. However, these

suggestions are too complex to implement hardware in the

practical applications. In recent years, many researchers have

been successfully implementing the wavelet neural network

(WNN) with its fast learning and wavelet decomposition

capability to control time-varying, nonlinear systems. The

time–frequency localization properties of wavelet function

and more efciently learning capability are the superiority

of the WNN over neural network (NN) [8–11]. Therefore,

* Duc-Nam Nguyen

[email protected]

Thanh-Quyen Ngo

[email protected]

Mien-Ka Duong

[email protected]

D. C. Pham

[email protected]

1 Faculty of Electrical Engineering Technology, Industrial

University of Ho Chi Minh City, Ho Chi Minh City, Vietnam

2 Division of Computational Mechatronics, Institute

for Computation Science, Ton Duc Thang University,

Ho Chi Minh City, Vietnam

3 Faculty of Electrical & Electronics Engineering, Ton Duc

Thang University, Ho Chi Minh City, Vietnam