Human-machine Systems Human-machine Systems

ions to describe a system have been proposed: c pltvsicalforni gives information about shape, location, weight, colour, etc.; ~ physical function gives information about functions of (groups of) components; e generalised function describes material flows and properties of cooling sources, heat and water storage, etc.; @ abstract function describes global product and energy flows in a process; @ functional purpose this highest abstraction level lists the goals of the process, which are at least safety, production and quality related. The use of the system description and presentation at each level requires different knowledge and reasoning processes. The physical form and sometimes the physical function form are the ones that are most often in, for instance, operator control rooms [Piping Instrumentation (P&I) diagrams]. It is a challenge to use highe abstractions in operator support systems (Bisantz and Vicente, 1994; Lind, i ~8~). For instance, instead of supporting the operator with a P&I diagram illustrating a pump that has failed (abstraction level: physical form) one might consider providing information on the function of the sub-unit that is no longer guaranteed due to the failed pump, i.e. cooling function failed or cooling-water inventory system not working properly. Additional support may help the operator to predict the ° consequences of that failed function for higher abstract © 1999 SAGE Publications. All rights reserved. Not for commercial use or unauthorized distribution. at PENNSYLVANIA STATE UNIV on April 17, 2008 http://tim.sagepub.com Downloaded from 145 objectives of the plant, such as the quality of the products. The use of abstraction hierarchies to describe a system is not widely accepted. It has been presented as a useful tool for developing representations of complex work environments (Bisantz and Vicente, 1994) by using physical and functional models. So far only a few examples from the cognitive engineering community exist where it has been applied to complex system (Ali, 1998; Larsen, 1993). The abstraction hierarchy given by Rasmussen can be applied at different levels of detail of the system; subsystems have goals. A collection of sub-systems that form the system will have another goal. This is called the partwhole relation which may be found at each level of abstraction. Lind (1988, 1990) developed a formalism for functional modelling which is called multi-level now modelling (MFM). An example of MFM is given by Lind in his contribution to this special issue (Lind, 1 ~99), An approximate functional modelling (AFM) technique has been developed by Van Paassen and ~Vierin~a (1998) for the purpose of alarm handling and fault diagnosis. The r~,Fl~~ was later extended (Van Paassen and Wieringa, 1999) to include temporal reasoning which opens the possibility for fault diagnosis and early warning systems. Operator interactions with a system during supervisory control are mainly monitoring (for data acquisition) and intervention (executing control actions). The human-machine interface is used by the operator to communicate with the system. The other supervisor control tasks, previously listed, relate to hypothesised, internal cognitive functions (Fig 4). Particularly during fault diagnosis and planning of an execution, cognitive functions of the operator are used (deduction and induction, mediumand long-term memory, system knowledge, etc.). Therefore, some knowledge about human cognitive functioning is needed in order to predict and understand human operator supervisory control behaviour, for instance, for the design of alarm handling philosophies and for the analysis of incidents and accidents due to possible human errors. Hollnagel (1999) illustrates the emerging need for cognitive engineering to model human operator behaviour after decades of modelling based on control engineering and information processing. Sheridan (1999) discusses the requirements for operator supervisory control of transportation systems, such as aircraft, rail and highways vehicles. Full automation of these systems in the near future requires radical change which is technologically, economically and socially not feasible. However, a clear trend towards the use of more computerised transportation systems can be seen. In order to use improved automation levels, an active debate should take place and empirically obtained performance and reliability indices should be used. 3. Understanding the control of complex