Invited paper (13g) to the 5th National Convention of Industrial Engineers, Quimica, Barcelona, 10th-12th March 1975.

S F Bush, ICI Europa.

An addendum – “A necessary condition for effective control in business systems” – was added in January 1979.  

Summary

Control of processes is conducted on a number of levels, from control of plant to the planning of projects. These levels are distinguished principally by their different working time-scales. The main means of dealing with control problems in the process itself has been to design them out. In situations where process and commercial time-scales are too long to rely for control on feedback alone, on-line computers have a unique role in predictive control. Current examples are taken to illustrate these points.

Introduction

The object of this paper is to expound basic control ideas relevant to the design and operation of chemical processes. No attempt has been made to include detailed performance figures of specific applications as would be required by control system designers. It is hoped, nevertheless, that even for them the presentation of principles may be relevant to the matching of the equipment specifications to the characteristics of the many different processes and industrial systems encountered in practice.

The control of processes can be seen narrowly as the provision of equipment for essentially regulation and scheduling, or very broadly as the design of complete business systems (including processes and managerial systems) to optimise some criterion (usually financial) over a period of time.

While the first view is too narrow to obtain full practical benefit from the available control ideas, these ideas do not as yet comprehend enough of the practical factors of process operations, factory constraints and commercial realities to make more than a relatively small contribution to the engineering and management skills deployed in design, and running a complex of factory processes.

Paper to the Institute of Mechanical Engineers Conference, Control in Process Engineering, 16th-17th May 1972.

S F Bush, ICI Europa.

Summary

Generally, the application of automatic control theory to chemical processes has not so far led to the same dramatic improvements as in electrical and mechanical systems. The paper seeks to determine the reason for this and traces the essential developments which seem to bear on this fact. It is found that in its approach to the control of chemical processes, control theory has been implicitly limited to working outside what may be termed the currency of a process. If this view is accepted, two new possibilities arise for the development of control theory and for its fruitful application: firstly, in the physico-chemical design of a process and secondly, in the design of computer control.

Thesis submitted for the degree of Doctor of Philosophy in the University of Cambridge, May 1965.

S F Bush

Summary

The subject of this thesis is the simplification and computation of mathematical representations, or models, of distributed chemical engineering systems in which there is flow and interaction of one or more quantities such as heat and mass. Counterflow heat-exchangers, regenerators, tubular reactors and packed distillation columns are examples of main types for which simplified representations have been obtained through the application of one of two methods. Among those systems for which one or other of the two methods is suitable, the choice of method is found to depend basically on the number of process transfer units in each stream.

One method, the diffusion representation, is based on a method applied by G I Taylor to the problem of solute dispersion in a capillary tube. In its present application the method replaces the set of partial differential equations, one for each stream, by a single parabolic partial differential equation. Such a transformation entails a certain error, mainly in the early part of the transient response, which is smallest for systems with the largest number of transfer units. If the error is acceptable, the transformation greatly simplifies subsequent numerical analysis.

The second method is an iterative method and is suitable for systems with a small number of exchange units. In contrast to the first method, it isolates the wave-like characteristics of exchange flow systems. In favourable cases it provides simple but accurate representations in either transfer function form or explicitly in the time domain. The iterations are shown to converge for a general class of flow system and form the basis of a numerical procedure for integrating the sets of first order hyperbolic differential equations which describe these systems.

The simplified representations obtained by application of the methods are compared with the standard representations by computing the step responses of both representations. The criteria of applicability which are proposed for the two methods are obtained by numerical experiments. Tables are presented comparing the numberical methods used in computing both simplified and standard models.

Thesis in partial fulfilment of the requirements for the degree of Master of Science, Massachusetts Institute of Technology, USA

S F Bush

Summary

The analysis and control of a continuous process can be approached in two ways. One method analyses the system as closely as possible to obtain equations which describe the process in detail. The second method seeks to find relations between outputs and inputs by statistical correlation methods without enquiring closely about the details of the process. This thesis deals with the use of the first method on the control of a chemical process producing hydrogen from steam and carbon monoxide.

Generally, the first method yields non-linear, partial differential, simultaneous equations. The main burden of this thesis is to obtain approximate solutions of these equations and to obtain them in a form suitable for controller design. A method for getting non-linear solutions has been worked out. The results obtained from this method have been compared with those obtained from considering small perturbations of the dependent variables about their steady state values. Solutions for the perturbations were found in two ways: firstly by an integral series technique, previously applied to the equations of a heat exchanger and extended here to solving linear equations with variable coefficients, and secondly by a lumping procedure which reduces the equations to ordinary differential equations. The result of all these methods are compared with each other and with numerical solutions and there is a short note on the methods which failed.

The reactor is controlled by altering the flow of reactor gas to a heat exchanger heating the incoming gases. In this case, the control problem is not difficult, once the analysis has been done. In this process, the transfer function relating output to input disturbances is easily derived and several means of control suggest themselves.

Finally, there are tentative suggestions for extending the non-linear solutions and for obtaining some precise criteria for deciding which method of analysis is likely to be the best in any given case.

Project report for Part II of the Engineering Tripos, Cambridge University, 1960.

S F Bush

Summary

This is a report on the analysis, design and building of a servo-mechanism to control the movement of a uniform steel rod. The servo-mechanism is required to maintain the rod in a vertical position by moving it at the bottom. The bottom of the rod is fixed to a radial arm by a ball-bearing and is allowed only to move perpendicularly to the arm.

The report has been written with the intention of recording, more or less in order, the things which were done in all three stages of the project. With this in mind, those parts of the analysis which were central to the problems presented are included in the main text, while the subsidiary parts, and an account of abortive attempts to solve problems, are contained in appendices.

This report is written to illustrate the methods of dealing with the problems as they arose in the course of proceeding from the first analysis to the final building of a rather unusual system; it is not intended to be a concise account of the final design. An alternative design, using a relay, and basically simpler than the one actually used is, however, given towards the end. The report ends with a short conclusion which surveys the whole project in retrospect and in particular considers the role of the theoretical analysis.