Introduction to Fuzzy Logic

In the early 1990s, scientists at the National Center for Atmospheric Research were asked to help the Federal Aviation Administration objectively determine which flight service stations throughout the United States handled the most hazardous weather conditions, and hence should be spared from congressional budget cuts. They had access to 15 years of meteorological data from each location, including winds, temperature, fog, rain, and snow at 1-min intervals, as well as information about the air traffic density and a number of other factors. However, the level of aviation hazard was not indicated by any single statistic, but by the nature, frequency and duration of the conditions and their combinations. How could the stations be ranked in a reasonable, objective way?

The scientists began by surveying a group of subject domain experts — pilots, meteorologists, and airline dispatchers — quizzing them on what factors, or combinations of factors, they considered most dangerous. Using the results of these surveys along with the repository of historical data, they then computed a hazard score for each flight service station, and ranked them. When the final report was presented to the group of experts, they opened it and began to laugh — everyone agreed that the stations with the most hazardous weather were at the top of the list. Unwittingly, the NCAR scientists had created a fuzzy logic algorithm, efficiently encoding the experts' knowledge in a set of rules that reproduced their approach to assessing the level of hazard presented by each unique set of weather conditions.