K.J. Burnham

Logical covariance matrix reset in self-tuning control

Abstract The paper investigates the application of a state-space self-tuning pole-placement strategy to a non-linear hydraulic test rig designed specifically for evaluating the effectiveness of self-tuning control schemes. The test rig is representative of a wide range of industrial systems exhibiting both non-linear characteristics and fast dynamic behaviour, such factors, until recently, prohibiting the effective use of linear self-tuning schemes. In recognition of this shortfall, a novel technique for enhancing the covariance matrix update stage within the parameter estimation algorithm of a self-tuning controller is proposed. The technique, termed logical covariance matrix reset, requires only a minimum of computational overhead and is shown to give rise to a significantly improved parameter tracking ability by making more effective use of potential information available to the estimation algorithm at set-point change. Results of simulation studies involving the application of the logical resetting technique to the hydraulic test rig are presented when the rig is subjected to slow variation in feed flow and sudden variation in load.

A reduced order self-tuning controller

As industrial users attempt to control systems exhibiting fast dynamic response characteristics, the demand for higher sampling rates and hence rapid controller calculations is becoming more apparent. This problem is compounded when considering the application of self-tuning control to non-linear systems exhibiting such characteristics, where, in order to maintain integrity of the self-tuner, enhanced estimation algorithms with increased computational complexities are required. In this paper a reduced order state-space self-tuning controller is proposed and its performance is compared by simulation studies to that of a standard implicit delay state-space self-tuning scheme. Whilst the same control objectives are achieved in a compatible manner it is shown that the reduced order scheme is significantly less computationally intensive and, as such, reduces the effects of computational delay in the overall closed-loop self-tuning scheme.

Industrial Applications of Self-tuning Control

Abstract The use of self-tuning control is now becoming well established in U.K. industry as witnessed by the availability of commercial self and auto-tuning controllers. As a consequence there is much on-going research activity in the refining and tailoring of algorithms in order to accommodate effective implementation on specific industrial applications. The paper considers three potential industrial systems, each exhibiting non-linear and/or fast dynamic behaviour. These are: hydraulic servo system, dynamometer torque loop and a heater-fan system. Drawing from the results of the three investigations, relative advantages of adopting features such as logical reset action, switched controller structure and bilinear model structures are discussed.

Identification of dynamic nonlinear polynomial models in the errors-in-variables framework

Abstract An approach for the identification of a class of discrete-time nonlinear polynomial single-input single-output dynamic errors-in-variables system models in the case of white input noise and white output noise is proposed. The algorithm is constructed within the extended bias compensated least squares framework. The separable nonlinear least squares method is subsequently utilised to determine the model parameters together with the input and output noise variances. A numerical Monte-Carlo simulation demonstrates the robustness of the proposed approach with regard to noise.

DESIGN OF SELF-TUNING CONTROLLERS FOR INDUSTRIAL PLANT

Abstract Self-tuning control is now rapidly maturing to provide a realistic option for the control of a wide range of industrial systems. This is particularly evident in the field of process control where system time constants are relatively slow and assumptions on slowly varying plant dynamics permits the use of such adaptive schemes. However, the application of self-tuning techniques to systems exhibiting fast dynamic behaviour and/or severe non-linear characteristics appears to be somewhat limited and it is this, coupled with recent advances in computer technology, that has provided the impetus for much interest and on-going research. By adopting a case study approach, the paper gives detailed consideration to the application of self-tuning control to an engine test cell which exhibits fast dynamic behaviour and to a high temperature gas fired heat treatment process which exhibits severe plant non-linearities. Drawing from these two industrial applications, some of the problems encountered when implementing standard self-tuning control schemes are highlighted and the need for the tailoring and refining of the algorithms for specific applications is emphasised.

Frisch scheme based identification of dynamic errors-in-variables bilinear systems

Abstract An approach for the identification of dynamic single-input single-output bilinear discrete-time system models within the errors-in-variables framework for the case of white input and output noise sequences is presented. The method is based on the combination of the instrumental variables technique and the Frisch scheme. Whilst the instrumental variables technique is utilised to avoid correlations of the noise sequences corresponding to the bilinear part of the noise covariance matrix, the Frisch scheme allows to obtain, subsequently, the model parameters together with the variances of the input and output noise sequences. The concept gives rise to two novel algorithms, which are compared with other errors-in-variables approaches via a Monte-Carlo simulation.

A fuzzy-logic approach to industrial control problems

Increasing demands for improved profitability and product quality, together with a growing awareness of the effects of industrial wastage on the environment, is forcing manufacturers to closely examine their process operations. As a consequence there is currently significant research and development activity aimed at improving control system strategies in a variety of industrial sectors. Recent years have witnessed renewed interest in fuzzy logic and rule-based control strategies and, by considering two illustrative industrial case studies, this paper highlights some of the potential advantages.

Design of Self-Tuning Controllers for Industrial Plant

Abstract Self-tuning control is now rapidly maturing to provide a realistic option for the control of a wide range of industrial systems. This is particularly evident in the field of proccss control where system lime constants arc relatively slow and assumptions on slowly varying plant dynamics permits the use of such adaptive schemes. However, the application of self-tuning techniques to systems exhibiting fast dynamic behaviour and/or severe non-linear characteristics appears to be somewhat limited and it is this, coupled with recent advances in computer technology, that has provided the impetus for much interest and on-going research. By adopting a ease study approach, the paper gives detailed consideration to the application of self-tuning control to an engine test cell which exhibits fast dynamic behaviour and to a high temperature gas fired heat treatment proccss which exhibits severe plant non-linearities. Drawing from these two industrial applications, some of the problems encountered when implementing standard self-tuning control schemes arc highlighted and the need for the tailoring and refining of the algorithms for specific applications is emphasised.

Filtered and Predicted States for Discrete-Time Adaptive Control

Abstract The paper considers the adaptive control of discrete-time, noise corrupted, systems whose characteristics can be represented by means of an ARMA model, this being posed within a canonical state-space framework. The effects of using different state vector forms, on self-tuning controller performance, are investigated with regard to filtered and predicted state representations when applied to various system types. It is shown how only the optimal observer for the state-space filter model is truly optimal in the sense of being a constituent part of an optimal controller. The important aspect of these results is that the state-space prediction model, also known as the innovations model, is that which is widely used in practice, despite the fact that overall optimality conditions cannot be obtained with a state-feedback controller based on this state-space model. The suggestion is made that the state space filter model, also known as the noise-free measurement model, is a much more viable alternative.