Aviation Visual Crack Measurement

摘要

This paper presents results from a NASA/Northwest Airline sponsored study designed to determine airline inspector reliability for finding cracks visually. A large data set from Northwest Airlines was analyzed by the University of Idaho to determine the reliability of the inspectors. This process is discussed in Ostrom and Wilhelmsen (ref. 1). The crack lengths in the data from Northwest Airlines were estimated by the inspectors. We needed to determine how well the inspectors actually estimated these lengths visually. We needed to do this because it would impact the estimated probability of detection (POD) for the cracks. For instance, if inspectors always underestimated the crack lengths, then the POD would be less than what would be predicted from the actual data. If the inspectors overestimated the crack lengths the POD would be more than what would be predicted from the actual data. We also wanted to determine if part type and color of the part would impact visual line length estimates. A visual line length measurement study was developed. The study was designed with three different parts, two colors, and 16 line lengths: 1/8 to 2 inches (the range of cracks we primarily saw in the data). There were a total of 96 treatments. A pilot study was done with 7 students at the University of Idaho in the early part of January 2002 to test the tool. We traveled to Northwest Airlines in Minneapolis in mid-January of 2002 and collected data from a total of 20 inspectors. They represented all three shifts (day, swing and midnight). Very preliminary results show that inspectors primarily overestimate line length, indicating they are more reliable than predicted by the Northwest Airline data alone.