A Fundamental Study on Dynamical Characteristics of Face Graph for Sensing of Plant Abnormality

This paper deals with dynamical characteristics of face graph, which is drawn by using mathematical formulas containing 21 parameters normalized within [0,1]. Face graph is a new method as an integrated display of multidimensional time series data such as plant states for an improved man-machine communication. Dynamical sensitivity expressing a degree of easiness of distinguishing a dynamical change in the face graph parameters like eye size, eye openness, nose length, mouth shape and so on, and learning effects with face graph were experimentally examined. It is clarified that 1) dynamical and statical sensitivity on face graph parameters showed a similar tendency, and 2) low dynamical sensitivity showed higher learning effects. For dynamical change in face graph's expressions, such as laughter, anger, sadness, smiling, surprise and suspicion, recognition characteristics analysis experiments, analysis of learning effects and of preliminary knowledge effects on face graph were carried out. Exeriment revealed that the anger showed the highest degree in recognition; the laughter and smiling showed the highest learning effect; and the laughter, smiling and surprise showed a high value for preliminary knowledge effect. An application of this dynamic face graph to a display of 4-dimensional time series data obtained from a heat exchanger plant model clarified that even a slight change in plant state is easily recognized through a dynamic display of plant state on face graph's expression. Keyword: face graph, dynamic display, visual sensitivity, recognition, application experiment 1. ま え が き