In complex and high-risk work conditions, especially such as in nuclear power plants, human understanding of the plant is highly cognitive and thus largely dependent on the effectiveness of the man-machine interface system. In order to provide more effective and reliable operating conditions for future nuclear power plants, developing more credible and easy to use evaluation methods will afford great help in designing interface systems in a more efficient manner.
In this study, in order to analyze the human-machine interactions, I propose the Human-processor Communication(HPC) model which is based on the information flow concept. It identifies the information flow around a human-processor. Information flow has two aspects: appearance and content. Based on the HPC model, I propose two kinds of measures for evaluating a user interface from the viewpoint of these two aspects of information flow. They measure the communicative complexity of each aspect.
In this study, for the evaluation of the aspect of appearance, I propose three complexity measures: Operation Complexity, Transition Complexity, and Screen Complexity. Each one of these measures has its own physical meaning. Two experiments carried out in this work support the utility of these measures. The result of the quiz game experiment shows that as the complexity of task context increases, the usage of the interface system becomes more complex. The experimental results of the three example systems(digital view, LDP style view and hierarchy view) show the utility of the proposed complexity measures.
In this study, for the evaluation of the aspect of content, I propose the degree of informational coincidence, R (K, P) as a measure for the usefulness of an alarm-processing system. It is designed to perform user-oriented evaluation based on the informational entropy concept. It will be especially useful inearly design phase because designers can estimate the usefulness of an alarm system by short calculations instead of costly operator-based tests. The result of the experiments for validating the proposed measure shows that as the problem contains larger R (K, P), the correctness of identifying process deviations becomes lower. I also propose a procedure based on the analytic hierarchy process (AHP) for evaluating alarm-processing systems with regard to integrating a series of deviations. It provides a relative ranking among alternative alarm systems by integrating the evaluation results on the usefulness of identifying deviations. The conventional AHP determines weights using a subjective pairwise comparison between each pair of criteria. In this study, I reduce the number of pairwise comparisons by introducing the measure, R (K, P). In order to show the validityof the proposed integrating procedure, I present an exemplary application. Generally, this AHP-based evaluation procedure successfully integrates a series of evaluation results for deviations and clearly shows relative ranks. More investigation including the enlargement on various deviations will provide a valuable guideline for the alarm system evaluation.
