Decision-Analytic Approaches for the Use of Leading Indicators in Risk Management.* V. M. Bier and Y. S. Huang, University of Wisconsin-Madison, 1513 University Avenue, Madison, WI 53706
The U.S. nuclear power industry is frequently concerned with evaluating trending data for use in risk management. For example, aging of plant components has become a significant concern in the face of nuclear plant license renewal. Trend analysis is also used in evaluating the significance of performance indicators. However, data will often be too sparse to permit reliable determination of trends over a reasonable time frame. In fact, due to the severity of nuclear plant accidents, we will generally not want to wait until we have high confidence that an observed trend is real before taking risk management actions. Therefore, important risk management decisions will often need to be taken in the face of substantial uncertainties. In this paper, decisions analysis is used to evaluate three candidate actions: status quo (i.e., do nothing); risk reduction (e.g., major component replacement or refurbishment activities); and data collection (e.g., root cause analysis, watchful waiting). The conditions under which each of these actions will be optimal are outlined. Factors affecting the optimal choice include: the costs of the various activities; the cost of an accident; the length of the time horizon; the, initial component failure rate; the conditional probability of an accident given a component failure; the nature of the trend (e.g., exponential or linear); the steepness of the trend; the extent of uncertainty about the trend; and the effectiveness of data collection in reducing that uncertainty. A representative example will be presented.
*Work supported by the U.S. Nuclear Regulatory Commission under Grant Number NRC-04-93-083.