Relating aviation service difficulty reports to accident data for safety trend prediction

Commercial aviation fatalities are predicted (Fulwood et al., 'Relating aviation service difficulty reports to accident data for safety trend prediction', BNL Tech. Rpt 63018, March 1996 ¹) using linear superposition of the time-dependent spectra of key aircraft systems¹ difficulties reporting in FAA's Service Difficulty Reports (SDR (DOT, FAA, 'Flight standards service difficulty program', Order 8010.2, Feb. 22, 1978, reissued 4:5/14/81 ²)) data base. The fitting coefficients are found by a linear regression model (referred to as 'the model') to FAA's Accident Incident Data System (AIDS (DOT/FAA, 'Aviation standards accident/incident data system-AIDS', Users Guide VS ASAS-D-335, July 1982 ³)) covering 5.5 years beginning January 1990. The model was tested by dividing the data approximately in half, using the first half to calibrate the model for prediction of the second half. A second test did the opposite. A third test used the first 60 months of data to predict the following 6 months. These tests (Fig. 5) showed good agreement between the model and AIDS data. The deficiency frequency of ATA (Aircraft Transportation Association) systems is reported (Table 2). Third-order fitting of the AIDS data was also used for prediction. All methods are compared in Fig. 4. The model was deemed superior because it reflects inspections and may be updated with SDR data from the Internet. The magnitude of the model's fitting coefficients indicate systems importances to the results.