Neural Network Forecasting of Service Problems for Aircraft Structural Component Groupings

Abstract

The Federal Aviation Administration (FAA) in the United States is responsible for regulating aircraft traffic and safety. A significant increase in domestic air traffic coupled with an aging population of aircraft has led the FAA to initiate new aircraft safety research efforts. These efforts are intended to provide the FAA's aviation safety inspectors (ASIs) with the means to evaluate and to control appropriate surveillance levels for aircraft operators. One of the FAA's databases is the service difficulty report (SDR) system. It provides FAA inspectors with reliability and airworthiness statistical data necessary for planning, directing, controlling, and evaluating certain assigned safety and maintenance programs. Neural network forecasting models are developed that predict the number of SDRs for aircraft structural component groupings, referred to as Air Transportation Association chapters. Data are used from two specific operators with homogeneous fleets, that is, same aircraft make. It is an extension of previous SDR forecasting research in that it stratifies forecasts by structural component groupings.