The effect of information types on diagnostic strategies in the information aid

https://doi.org/10.1016/j.ress.2005.11.061Get rights and content

Abstract

Through experiments, this paper investigates the compatibility of information types with the diagnostic strategy in information aid. Compatibility with operator strategies is an important requirement for information aiding systems in nuclear power plants (NPPs). This paper used three typical types of information aids for MCR operators to investigate the effect of the aids on diagnostic strategies: “alarm (A),” “hypothesis on faults (H),” and “hypothesis on faults and expected symptoms (HS).” The experimental results indicate that the effectiveness of information aid types can vary, dependent on the strategies subjects employ. The results also show that the HS aid improved the diagnosis performance in the hypothesis-and-test strategy.

Introduction

In nuclear power plants (NPPs), the operator's actions must always be based on identification of the operational state of the system. To identify the state means to give it a name, that is, to label it in terms that will refer either to the functional state of the system, to the cause of this functional state, or, directly, to the related control action [1]. Diagnosis is a prerequisite task for initiating operating procedures that describe predefined steps and is usually carried out in main control rooms (MCRs) in order to correct the faults. In NPPs, identifying the cause of the plant state takes priority over producing compensatory actions, since the operating procedure already contains the actions that should be performed if an operator selects the correct procedure. Because operators know the appropriate response for a given causality well, most difficulties in NPP operation result from an inability to identify the nature of the problem [2]. As witnessed in the TMI accident [3], [4], operator misdiagnoses and incorrect situation awareness may contribute to the propagation or occurrence of accidents.

As the capabilities of modern computers in processing and information presentation have increased, modern computer techniques have been introduced into the design of advanced MCRs of NPPs [5]. The advance in the technology made it possible to improve operator diagnosis and situation awareness. The approaches can be categorized into direct and indirect supports. The indirect support is to improve the graphic displays of MCRs. This approach involves the improvement of integrated graphic displays, configural displays [6], and ecological interface design [7], [8], [9], [10]. The direct support is to develop decision support systems. These include intelligent advisors, alarm systems, computer-based procedures, fault diagnostic systems, and computerized operator support systems (COSSs), which are based on expert systems or knowledge-based systems. The main purpose of these approaches is information aiding, which supplies useful information to the human problem solving [11]. The computer may enhance the human diagnostic performance by supplying relevant information.

Strategy in human decision-making is defined as a sequence of mental and effector (action on the environment) operations used to transform an initial state of knowledge into a final goal state of knowledge [12]. When a diagnosis task is regarded as what an operator should do, diagnostic strategies can be viewed as the way the operator actually performs the diagnosis task. The informational need of a diagnostician may change according to the strategy that he/she currently employs. Therefore, it is necessary to meet the unique needs of each strategy to create computer-based support.

This paper investigates the effects of information aid types on the diagnostic strategies through experiments. The main purpose of this paper is to identify which type of information aid is useful for a specific strategy. The experimental results show statistically that the hypothesis that the effects of information aid types on diagnosis performance would differ according to the strategies that subjects employ is plausible. This paper also provides discussions on the application of the results to actual plants. Four different types of information aids that are representative features of direct support systems for MCR operators were used for the experiments: (1) no aid, (2) alarm, (3) hypothesis with certainty factor, and (4) hypothesis with certainty factor+symptom patterns. The diagnostic strategy taxonomy proposed by Rasmussen [1], [13] was used for strategy categorization, in particular, the symptomatic search.

Section snippets

Diagnostic strategy

Payne et al. defined decision strategies in human decision-making as a sequence of mental and effector (action on the environment) operations used to transform an initial state of knowledge into a final goal state of knowledge [12]. From the perspective of a cognitive work analysis [14], [15], if tasks or control tasks are the goals that need to be achieved, independently of how they are to be achieved or by whom, strategies can be defined as the generative mechanisms by which a particular task

Objective

This paper aims to investigate the impact of information aiding types on diagnostic strategies through experiment. The following hypothesis and interrogation are tested in the experiments.

  • (1)

    The effect of information aiding types would be different according to the diagnostic strategies. That is, an information type would be effective for a diagnostic strategy and not for other strategies.

  • (2)

    Which type of information is useful as an aid for a specific strategy?

In addition, the change in selection of

Analysis of the results

The results of the work indicate that the first hypothesis in Section 3.1, that the effects of information aid types would vary, dependent on the strategies subjects employ, was verified. For the second hypothesis, in summary, for the pattern recognition strategy (Problem 7), there was no significant difference between the three aiding types (that is, A, H, and HS aids) and N aid. For the decision table strategy (Problems 1, 4, and 5), there was no significant difference between the four aiding

Conclusions

This paper presented a strategy-based evaluation of the effects of information aiding types on the diagnosis performance. Four types of information aiding elicited from typical direct support systems of NPPs were evaluated. The effects of these information aiding types were investigated through experiments. The experimental results were then analyzed based on subject strategies. The experiments show that the effects of information aiding types can vary according to the strategies. The results

Acknowledgments

This research was supported by “The National Research Laboratory Program” of MOST (Ministry of Science and Technology), Korea.

References (46)

  • J. Bernard

    Issues regarding the design and acceptance of intelligent support systems for reactor operators

    IEEE Trans Nucl Sci

    (1999)
  • Kemeny J. The need for change: the legacy of TMI. Report of the President's Commission on the Accident at Three Miles...
  • J. Reason

    Human error

    (1990)
  • N. Moray

    Human factors in process control

  • K.J. Vicente et al.

    Ecological interface design: theoretical foundation

    IEEE Trans Syst Man Cybern

    (1992)
  • N. Dinadis et al.

    Ecological interface design for a power plant feedwater subsystem

    IEEE Trans Nucl Sci

    (1996)
  • W.C. Yoon et al.

    Deep-reasoning fault diagnosis: an aid and a model

    IEEE Trans Syst Man Cybern

    (1988)
  • J.W. Payne et al.

    The adaptive decision maker

    (1993)
  • J. Rasmussen

    Information processing and human–machine interaction

    (1986)
  • K.J. Vicente

    Cognitive work analysis

    (1999)
  • J. Rasmussen et al.

    Cognitive systems engineering

    (1994)
  • J. Rasmussen et al.

    Mental procedures in real-life tasks: a case study of electronic trouble shooting

    Ergonomics

    (1974)
  • J.W. Payne et al.

    Risky choice: an examination of information acquisition behavior

    Mem Cognit

    (1978)
  • Cited by (0)

    View full text