A model and experimental analysis of pilot decision making behaviour for various approach conditions

This report deals with an investigation of pilots' failure detection performance in a realistic moving-base flight simulator. The task considered was the monitoring of stabilized automatic approaches. Abnormal conditions (excessive windshears or system failures) were to be detected on the basis of unacceptable deviations of displayed variables from their normal position. Four experienced airline pilots participated in the experiment. The independent variables were: failure type (windshear or system failure), the axis in which the failure occured (longitudinal or lateral), and the prior probability of failure occurrence (0.2 or 0.8). The dependent variables were detection times and the various display deviations at the moment of detection. Independently of the experimental program a theoretical analysis was carried out, based on a model of the human observer and decision maker which is formulated in terms of linear estimation and classical sequential decision theory. In the report the experimental results are compared with the corresponding model predictions. In addition, attention is devoted to the extrapolation of the results for the detection of relative less frequently occurring failures.