Site: Central Research Institute of the Electric Power Industry (CRIEPI)
Human Factors Research Center
2-11-1 Iwado Kita, Komae-shi
Tokyo 201, Japan

Date Visited: May 25, 1995

Report Author: R. E. Kraut



S. Chipman
R. E. Kraut
T. Sheridan


Seiichi Yoshimura
Kunihide Sasou
Naoko Hasegawa
Brian Landberg


The Central Research Institute of the Electric Power Industry (CRIEPI) was founded in 1951. Its charter is to conduct research on ways to improve the generation of electricity, especially for the nuclear power industry. The Human Factors Research Center was begun in July 1987, obviously influenced by the U.S. Three Mile Island accident and the obvious great interest of America and Europe at that time in putting additional resources into research on training, procedures, control room design, and other human factors considerations. Its goal is the reduction of human errors in the nuclear power industry.

CRIEPI is spread over several locations, including Komae, Abiko, Yokosuka, and Agaki. The JTEC team visited the Komae facility in Tokyo, specifically the Human Factors Research Center. Researchers with backgrounds in nuclear engineering and psychology conduct physiological, cognitive modeling, and survey research on operator behavior in nuclear power plants and the factors that lead to errors in performance. Some of their work is conducted in collaboration with the Electric Power Research Institute in the United States. The researchers compile statistics on human factors around the world and analyze the data for its relevance to human errors in the electrical industry. Their output includes devices for measuring the physiological and psychological workload of human operators, procedural manuals for analyzing and reporting on human errors, guidelines and systems for predicting team performance, and tools for maintenance workers, as well as research reports. The Human Factors Research Center does not conduct research on human-computer interaction per se. However, the research center is interested in human performance when operating computer-assisted power plants and in using computer simulation to model operator-plant interaction.


Operator-Power Plant Simulator

CRIEPI conducts programs relevant to human-computer interaction or person-machine interaction more generally. Of particular interest was an experimental simulator, where teams of three operators were put through simulated emergencies and video/audio data were recorded. Researchers analyzed these in fine-grain detail and built a model of mental activity needed to respond to emergencies. The system is called SYBORG -- Simulation System for the Behavior of an Operating Group (Sasou et al. 1995). The simulation has three major subsystems: a power plant model, a human operator team behavior model, and a man-machine interface (MMI) model (see Fig. CRIEPI.1). The plant simulation models the power generation system, the controls, and the alarms in the plant. The operator model attempts to simulate both individual perception, cognition and behavior and group decision making as a team monitors and operates the plant and responds to alarms. The operator model receives information from the plant model and exchanges information to decide what team actions are necessary.

Fig. CRIEPI.1. Elements of the SYBORG system.

The operator has seven subsystems: an attention subsystem to model visual and auditory processing of plant alarms; a short-term memory subsystem; a long-term memory model that contains a knowledge base about the operation of power plants, the meaning of alarms, and appropriate behavior for handling them with reasoning; a thinking module that runs the mental model enabled by the long-term memory to make decisions on receiving plant alarms; an utterance module to communicate among team members; and an action module that allows the operators to move around a control room and push switches.

The researchers are comparing the sequence of actions taken by the simulated operator's crew on hearing various emergency alarms with the behavior of real nuclear engineers of varying experience levels as they run though laboratory exercises. While analysis is still ongoing, the researchers report substantial correspondence between predictions made by the computer simulation and the utterances and actions made by experimental teams.

Human Error Research

The CRIEPI group is also conducting more empirically oriented work for cataloging, systematizing, and analyzing individual and group errors in the operation of nuclear power plants. These efforts include a Human Error and Behavior Prediction System, which is a rule-based, causal model for predicting error. The model is of the form "Too much information is being supplied at the same time ~ leads to ~ Operators don't know what to attend to ~ leads to ~ Operators misjudge the cause of an alarm ~ leads to ~ Error in decision-making ~ leads to ~ A specific error action ~. Rules were generated by sending questionnaires to 3,500 experts with at least five years experience in plant operations and maintenance, asking them the extent to which various error sources identified from a literature review were associated with plant errors.

The researchers' analysis of the causes of errors in the nuclear power industry attributes includes a number related to the human-machine interface, including having poorly visible indicators for devices, too much information arriving at the same time, and feedback on the results of an operation arriving too late. Other categories of causes include internal attributes of the operators (e.g., too little experience), work condition (e.g., task is infrequently done), organizational factors (e.g., work responsibilities are not well established), and environmental factors (e.g., the workplace is too noisy).

The research resulted in a computer program for assessing the severity of sources of error.

The CRIEPI group has also constructed a Japanese version of the Human Performance Enhancement System, a technique for reporting on and diagnosing the root causes of error reports in power plants. This is an attempt to systematize the analysis and evolution the root causes of particular error episodes. The system provides a technique for collecting data, analyzing causes, and proposing countermeasures based on the Human Performance Enhancement Systems (INPO 1989).

Human Performance Evaluation

The Center has developed a wearable Human Performance Monitoring System for assessing the physical and physiological responses of plant operators as they go about their work. The system consists of position sensors, eye movement sensors, body movement sensors, and physiological function sensors. The system is designed to provide performance feedback in experimental scenarios involving the redesign of plant equipment and layout or work practices.


While the Human Factors Center is doing a wide range of work relevant to plant safety, the Operator-Power Plant Simulator is most relevant and interesting from the point of view of human-computer interaction. It is noteworthy because it extends techniques developed in the United States on cognitive models by adding the interaction among individuals in a team and by meshing the human model with a simulation of an external dynamic system (a power plant). Finally, it is using laudable research methods, by detailed comparisons of the computer simulations with experimental studies of real human teams.


Sasou, K., K. Takano, S. Yoshimura, K. Haroko, and M. Kitamura. 1995. Modeling and simulation of operator team behavior in nuclear power plants. HCI international '95. Tokyo.

Sasou, K., S. Yoshimura, K. Takano, S. Iwai, and J. Fujimoto. 1993. Conceptual design of a simulation model for team behavior. Fourth European conference on cognitive science approaches to process control. Frederiksborg, Denmark.

Takano, K., K. Sawayanagi, and T. Kabetani. 1994. System for analyzing and evaluating human-related nuclear power plant incidents: Development of a remedy-oriented analysis and evaluation procedure. Journal of Nuclear Science and Technology 31(9): 894-913.

Yoshimo, K., and K. Inoue. 1993. Practical development of human errors and Behavior Prediction System (HEBPS) from the viewpoint of psychological and statistical methods. In Safety and reliability assessment: An integrated approach, ed. P. Kafka and J. Wolf, 949-959. NY: Elsevier Science.

Institute of Nuclear Power Operators (INPO). 1989. Human performance enhancement system: Coordinator manual. INPO 86-016 (Rev. 02).

K. Mitsuhiro. N.d. Industrial changes in human consciousness. Apparently unpublished paper.

Published: March 1996; WTEC Hyper-Librarian