Manual control of slow response systems (extract)

 This paper presents the results from laboratory studies of controlling a slow response system.  

In the simple process studied (making a step change to the temperature of a water bath by varying the electric heating to it), there was a time lag of 2 minutes between making a control action and when it had had its full effect on the system output.

Other sections of the paper, not reproduced here, attempt to model the operator’s behaviour using mathematical equations, as was being done at that time for fast response systems such as flying.

This extract shows one result which I refer to, as it shows learners set a slow-response process into oscillation when they first try to control it.

This is relevant to manual take-over from automated systems, as this sort of control skill is only retained with frequent practice.

Manual control of slow response systems


by E. R. F. W Crossman and J. E. Cooke

Institute of Experimental Psychology, Oxford University

presented at International Congress on Human Factors in Electronics, Long Beach, California, 1962.

reprinted in E. Edwards and F. P. Lees (1974)  The Human Operator in Process Control.  Taylor & Francis, London.

Figure 2 - typical results

[time lag - 2 minutes.

dashed line - temperature.

stepped line - heater setting]

[from page 53]

In the L condition [time lag 2 minutes] (see Figure 2b) subjects experienced more difficulty in achieving stability.  Some managed to do so within the first trial (in 20 minutes or so) but others went on for five or six trials without managing to suppress the initial oscillation of temperature.  During this initial period, the temperature typically swung two or three degrees above and below the target in a cycle with a period of five to ten minutes, and the control settings also varied in the same way but in opposite phase.  With increasing experience, subjects reduced the amplitude of oscillation and also lengthened the period, finally achieving stability at about the same settings as were required in the S [short time lag] condition.  Highly practised subjects in the L condition could also manage with very few control changes [Figure 2c].

[from page 56]

3.2.  The effect of instructions

A further experiment was carried out to determine the value of different types of prior instruction on performance.  A group given a scientific account of the manner in which heat was transferred from heating coil to water resulting in a change of temperature, performed markedly worse than one told simply what voltage settings were likely to produce what temperatures and roughly how much lag was to be expected.  This result casts some doubt on the value of the widespread practice of instructing process plant operators in the elements of chemistry and physics.

4. Analysis of Control Behaviour 

We were not able to extract much information from subjects’ verbal accounts of their control behaviour or thought processes, except to suggest that there was very little conscious decision-making in connection with control changes.  Most of the remarks were of the type "Yes I think it’s about time it should go up now" and "I seem to find 82 is quite a good setting".  One or two subjects initially expected to find a lag in the system but they were not at all clear about what caused it or how long it might be expected to be.  Lacking direct information about the manner in which subjects make control decisions, we had to fall back on examining the detailed relationship between control behaviour and the current situation.

[Most of the people tested were Oxford university students, so not usually poor at expressing themselves verbally, though many were not science students.]

Access to other papers via Home page

© notes  2022 Lisanne Bainbridge


Popular posts from this blog

Ironies of Automation

Types of skill, and Rasmussen's SRK schema

Complex Processes Review : References