Skip to main content
main-content

Über dieses Buch

The series Advances in Industrial Control aims to report and encourage technology transfer in control engineering. The rapid development of control technology has an impact on all areas of the control discipline. New theory, new controllers, actuators, sensors, new industrial processes, computer methods, new applications, new philosophies ... , new challenges. Much of this development work resides in industrial reports, feasibility study papers and the reports of advanced collaborative projects. The series offers an opportunity for researchers to present an extended exposition of such new work in all aspects of industrial control for wider and rapid dissemination. Neural networks is one of those areas where an initial burst of enthusiasm and optimism leads to an explosion of papers in the journals and many presentations at conferences but it is only in the last decade that significant theoretical work on stability, convergence and robustness for the use of neural networks in control systems has been tackled. George Rovithakis and Manolis Christodoulou have been interested in these theoretical problems and in the practical aspects of neural network applications to industrial problems. This very welcome addition to the Advances in Industrial Control series provides a succinct report of their research. The neural network model at the core of their work is the Recurrent High Order Neural Network (RHONN) and a complete theoretical and simulation development is presented. Different readers will find different aspects of the development of interest. The last chapter of the monograph discusses the problem of manufacturing or production process scheduling.

Inhaltsverzeichnis

Frontmatter

Chapter 1. Introduction

Abstract
Man has two principal objectives in the scientific study of his environment: he wants to understand and to control. The two goals reinforce each other, since deeper understanding permits firmer control, and, on the other hand, systematic application of scientific theories inevitably generates new problems which require further investigation, and so on.
George A. Rovithakis, Manolis A. Christodoulou

Chapter 2. Identification of Dynamical Systems Using Recurrent High-Order Neural Networks

Abstract
The use of multilayer neural networks for pattern recognition and for modeling of “static” systems is currently well-known (see, for example, [1]). Given pairs of input-output data (which may be related by an unknown algebraic relation, a so-called “static” function) the network is trained to learn the particular input-output map. Theoretical work by several researchers, including Cybenko [16], and Funahashi [24], have proven that, even with one hidden layer, neural networks can approximate any continuous function uniformly over a compact domain, provided the network has a sufficient number of units, or neurons. Recently, interest has been increasing towards the usage of neural networks for modeling and identification of dynamical systems. These networks, which naturally involve dynamic elements in the form of feedback connections, are known as recurrent neural networks.
George A. Rovithakis, Manolis A. Christodoulou

Chapter 3. Indirect Adaptive Control

Abstract
This chapter is devoted to the development of indirect adaptive control techniques (based on RHONNs), for controlling nonlinear dynamical systems, with highly uncertain and possibly unknown nonlinearities.
George A. Rovithakis, Manolis A. Christodoulou

Chapter 4. Direct Adaptive Control

Abstract
This chapter is devoted to the development of direct adaptive neurocon-trollers for afflne in the control nonlinear dynamical systems possessing unknown nonlinearities. The recurrent high-order neural networks are used as models of the unknown plant, practically transforming the original unknown system into a RHONN model which is of known structure, but contains a number of unknown constant-value parameters, known as synaptic weights. When the RHONN model matches the unknown plant, we provide a comprehensive and rigorous analysis of the stability properties of the closed loop system. Convergence of the state to zero plus boundedness of all other signals in the closed loop is guaranteed without the need of parameter (weight) convergence, which is assured only if a sufficiency-of-excitation condition is satisfied. Moreover, certain sources of instability mechanisms, namely modeling errors, uncertainty in model order and external disturbances acting both additively and multiplicatively, are also considered. Modifications on the control and update laws are provided, to guarantee a certain robustness level.
George A. Rovithakis, Manolis A. Christodoulou

Chapter 5. Manufacturing Systems Scheduling

Abstract
Efficient management of the production process, in manufacturing systems is a vast subject with decisive impact on major economic entities, like productivity, competitiveness and viability. Production scheduling (i.e., organization and control of production operations in manufacturing systems), is essential for the efficient operation of a production environment. Production scheduling actually concerns the efficient allocation over time of resources for the manufacture of goods, involving decisions such as part release, routing, machine scheduling, set up times etc., with the objective of producing customers’ demands in a timely and economic fashion. Consequently, scheduling attempts to find a way to assign and sequence the use of these shared resources, such that the production constraints are satisfied and production costs are minimized.
George A. Rovithakis, Manolis A. Christodoulou

Chapter 6. Scheduling Using Rhonns: A Test Case

Abstract
In Chapter 5, the non-acyclic FMS scheduling problem was considered to be a control regulation problem, where system states (buffer levels), have to reach some prespecified production requirements, by means of control input commands. Based on a recurrent high-order neural-network model of the buffer states, an adaptive continuous-time neural-network controller was developed. Stable control and update laws guaranteeing system stability, boundedness of all signals in the closed-loop system and a uniform ultimate boundedness property of the control error were derived. Dispatching commands were issued by means of a discretization process of the continous control input. Furthermore, modeling errors and discretization effects were taken into account, thus rendering the controller robust and capable of driving system production to the required demand.
George A. Rovithakis, Manolis A. Christodoulou

Backmatter

Weitere Informationen