ABSTRACT
Little advancement has been made upon the model of alarm handling proposed by Lees (1974) over two decades ago. This model comprised three stages: detection (detecting the fault), diagnosis (identifying the cause of the fault), and correction (dealing with the fault). It appears to be very similar to the process model put forward by Rouse (1983) comprising: detection (the process of deciding that an event has occurred), diagnosis (the process of identifying the cause of an event), and compensation (the process of sustaining system operation). A similar model, comprising: detection (detection of the onset of a plant disturbance), diagnosis (diagnosing the particular disturbance from presented symptoms), and remedial actions (selecting and implementing the appropriate actions to mitigate the disturbance) was proposed by Marshall and Baker (1994) as an idealized three-stage decision model to describe how operators dealt with faults. There appears to be little to distinguish these three models apart from the idiosyncratic labeling of the last stage in all cases. Rouse (1983) offers an expanded version of the process model comprising three levels: recognition and classification (in which the problem is detected and assigned to a category), planning (whereby the problem solving approach is determined), execution and monitoring (the actual process of solving the problem). Arguably this is reducible to the original three-stage model, but rather more emphasis has been place upon the interpretation of the problem. What is not clear from any of the analyses presented with these models is whether they accurately reflect processes undertaken by the operator within the alarm handling task.
