Adaptive neuro-fuzzy PID controller for nonlinear drive systemby Piotr Derugo, Krzysztof Szabat

COMPEL

Text

Adaptive neuro-fuzzy PID controller for nonlinear drive system

Piotr Derugo and Krzysztof Szabat

Institute of Electrical Machines, Drives and Measurements,

Wroclaw University of Technology, Wroclaw, Poland

Abstract

Purpose – Various control structures and approaches are in use nowadays. Development of new ideas allows to obtain better quality in control of different industrial processes and hence better quality of products. As it may seem that everything in the classical systems has already been discovered, more and more research centres are tending to incorporate fuzzy or neural control systems. The purpose of this paper is to present an application of the adaptive neuro-fuzzy PID speed controller for a DC drive system with a complex nonlinear mechanical part.

Design/methodology/approach – The model of the driven object including such elements as nonlinear shaft with backlash and friction has been modelled using Matlab-Simulink software.

Afterwards experimental verification has been made using a dSPACE control card and experimental system with two DC motors connected with an elastic shaft.

Findings – The presented study shown that the adaptive controller is able to damp the torsional vibration effectively even for the wide range of the system nonlinearities. What is more the design approach for controllers design parameters has been described. Proposed approach is based on requested properties of system. Using proposed tuning scheme no detailed information about the object are needed.

Originality/value – This paper presents for the first time fully an PID adaptive neuro-fuzzy controller. The inputs are the weighted tracking error, error’s derivative and integrated error. What is more the adaptation algorithm consists of a model tracking error its derivative and integer. Also the proposed tuning algorithm in such a form is an original outcome.

Keywords Adaptive, Friction and backlash, Fuzzy controller, MRAS, Neuro-fuzzy,

Parameters design

Paper type Research paper

I. Introduction

Nowadays all industrial drive systems are required to possess the best possible control properties. Due to relatively cheap hardware, complicated control structures can be easily implemented in order to improve different criteria. One of the most desirable criteria is for the system to follow the reference signal with a very high accuracy and minimization of the transition phase. It is very important for such structures as robots, automatons and high-precision drives. In such applications the existing nonlinearities, structural uncertainties, parameter variations and vibrations affect the characteristics of the system (Amrane et al., 2013; Orłowska-Kowalska et al., 2009).

In order to provide the desired shape of the system states and taking into account the above-mentioned circumstances different control algorithms are developed.

It has been reported that a conventional control structure utilizing a simple PI

COMPEL: The International

Journal for Computation and

Mathematics in Electrical and

Electronic Engineering

Vol. 34 No. 3, 2015 pp. 792-807 ©EmeraldGroup Publishing Limited 0332-1649

DOI 10.1108/COMPEL-10-2014-0257

Received 10 October 2014

Revised 14 January 2015

Accepted 14 January 2015

The current issue and full text archive of this journal is available on Emerald Insight at: www.emeraldinsight.com/0332-1649.htm

This research work is supported by National Science Centre (Poland) under grant: Adaptive fuzzy control of the complex drive system with changeable parameters, UMO-2011/03/B/ST7/ 02517 (2012-2015). 792

COMPEL 34,3 controller cannot ensure a satisfactory performance of an electric drive with a vibration mode. A more effective approach is active modal damping, in which state variables associated with the vibrations are fed back together with the control structure (Szabat and Orłowska-Kowalska, 2007). In the literature more advanced control paradigms based on the application of nonlinear, adaptive or predictive control concepts can be found (Hace et al., 2005; Vašak and Perić, 2010). Among the advanced control frameworks are algorithms based on the fuzzy, adaptive theory (Szabat and

Dybkowski, 2010a, b; Szabat and Orłowska-Kowalska, 2007). Due to the nonlinear control characteristic of the object the fuzzy controllers can provide more flexible properties than classical ones.

In the literature, there is a number of studies showing the application of fuzzy systems in various industries including electrical drives. Most studies are based on the basic structures of neuro-fuzzy controller systems (Mamdani, TSK). In recent years, there have been various modifications to conventional neuro-fuzzy systems, such as the use of: sets of type II, recursive feedbacks, wavelet functions and additional compensators (Gökbulut et al., 2006; Wang et al., 2004; Wai and Chu, 2007; Xiaoli et al., 1997).

There exist studies about fuzzy PID controllers (Petrov et al., 2002), or different hybrids of PID with Fuzzy and Neural theory (Fallahi and Azadi, 2009; Soyguder et al., 2009). These modifications are designed to get better dynamic properties of the controlled process.

This paper presents the application of PID adaptive neuro-fuzzy controller (ANFC) for nonlinear drive system. Contrary tomost popular approaches (e.g. Szabat and Dybkowski, 2010a, b; Szabat and Orłowska-Kowalska, 2007), where the PI or PD type of fuzzy controllers are used in this study the utilization of the PID type fuzzy controller with three inputs is investigated. The existence of the additional part of the controller provides more flexibility in the control function modelling and ensures much better performances for such nonlinear control object. This paper, though, presents neuro-fuzzy adaptive PID controller which is not a gain scheduler for the classical controller but a direct N-F.

II. Object and control structure

In this study two mass DC drive system with elastic coupling has been modelled for simulation purposes using Matlab-Simulink software. Equations (1) and (2) describe the