On November 13th, Jaap van den Top gave a presentation to the Dutch section of the Institution of Railway Signal Engineers, on his thesis entitled “Process Control and Railway Safety”, on which he hopes to receive a doctorate from Delft Technical University on April 19th, 2010. The subject of his thesis is research into the phenomenon of Signals Passed at Danger in relation to the interaction between train drivers, expectation patterns, the Dutch Signal System and the role of the train and traffic controllers. His main argument revolves around the thesis that most accidents can be categorised as an uncontrolled transfer of energy (the train’s kinetic energy). Whereas signalling systems process and present information to drivers related to energy and place to control the energy in the system, process control information should relate to time and place (as exemplified by the diagrams in a timetable). Hence the signal system, in terms of cybernetics research done by Ashby in 1956, is a language that does not posses the ‘requisite variety’ in terms of the process control of the railway: signalling only conveys a message to the driver once deviations from the conflict-free timeline have grown such large that a conflict is about to happen. When combining aspects of human factors (skill-based, rule-based and knowledge-based behaviour) we recognise most driver actions are skills based, i.e. behavioural patterns based on expectations, which is fine as these take less time and mental effort to process.
However, when a train driver is presented with a situation which differs from his expectations, e.g. a short route into a platform rather than the longer one he is used to, the Dutch signal system cannot always convey that information to him and the driver does not recognise the ‘different scenario’ he is in. In such a case, the skill-based action no longer applies, but the information required to make a rule-based decision is lacking, even if the driver realised he needs to make one. Many accidents then in fact can be regarded as a difference between the plan the driver executes and the plan the traffic controller had in mind.
The same sort of error modelling theory was used to illustrate some poor examples of signal placing and sighting and show some examples of how other railways’ signal systems sometimes provide more appropriate solutions.
In an interesting aside Mr van den Top addressed the issue of the route knowledge which most railway require and sometimes takes on mythical proportions. But in fact this can never be allowed to be used as a substitute for proper signal placement and sighting. Furthermore, since it is based on engrained driver expectations, is difficult and costly to maintain and it is not ‘fail safe’, it should only be allowed to be supplemental rather than leading and is in itself a behavioural expectation pattern, giving rise to risks in unexpected scenarios.
In-cab signalling such as that in ETCS eliminates much of the ambiguity of the scenarios caused by signal placement and sighting, whilst obviously also providing brake and speed curve supervision and an option to add process control information, which, however, currently only receives little attention.
Finally, as a last remark, Jaap van den Top left us to ponder whether signal engineers should be required to be totally devoid of humour: Humour exploits mistaken expectation patterns, resulting from the use of logically correct but contentwise unexpected messages. A professional requirement of signal engineers could therefore be that they are totally devoid of a sense of humour, or the exact opposite requirement.