A methodology for predicting and mitigating loss of control incidents for general aviation aircraft

In comparison with other modes of transportation, aviation has earned a clear distinction as the safest mode of travel. In recent years aviation has also achieved steady improvement in the accident rates, further distinguishing the safety of aviation with respect to other transportation modes. When aviation accidents do occur, however, it has been found that the most likely cause of these accidents is loss of control (LOC). Annual analysis of accident data indicates that LOC is consistently the most common cause of aviation accidents and fatalities for commercial aircraft worldwide and the Federal Aviation Administration (FAA) identifies LOC as the most important safety concern for general aviation (GA) as well. Recent work to identify and mitigate LOC events has been largely successful in identifying the sequence of events that typically precedes a LOC incident. Using this knowledge, several proposals have been made to break this sequence through application of advanced techniques and methods to detect, mitigate, or recover from events that may lead to LOC. These methods often assume the presence of advanced vehicle systems, such as advanced avionic systems and automated aircraft control, which imply intended application to future aircraft systems. Many existing aircraft are not equipped with such systems, leaving a gap between existing aircraft capability and the proposed solutions to address LOC. This is particularly true for GA, where the average age of an active vehicle in the GA fleet is estimated by the FAA to be 40 years old, suggesting that the typical GA aircraft lack such advanced on-board systems. The objective of this dissertation is to develop a methodology which enables the identification and mitigation of LOC for a typical GA fixed wing aircraft. The methodology which is developed within this work seeks to satisfy this objective through a combination of three key components. First, as LOC is understood within the existing literature as a deviation of the aircraft from normal operation, an appropriately defined LOC envelope will enable the prediction of LOC onset. Then to monitor this envelope during flight all necessary states of the vehicle must also be either observed or estimated. As it is assumed that only data collected by personal electronic devices is available, unobserved aircraft states and control actions of the pilot must be estimated within the methodology using existing or developed techniques. Finally, the methodology will aid in recovery of the aircraft in the event of LOC through synthesis of LOC recovery strategies which would be communicated to a human pilot through aural cues. This proposed methodology is summarized as a method of Mitigation by Envelope Restriction for Loss-of-control INcidents (MERLIN). The dissertation presents tools for implementing each of these components and includes a set of methods for synthesizing a dynamic vehicle model for use alongside the method. The various aspects of this methodology are also tested through a series of experiments. First the primary sources of uncertainty which affect the LOC envelope estimation process are identified and studied, yielding quantification of the effects of this uncertainty on the envelopes and a strategy for compensation. Secondly the expected error of the estimation of flight states is analyzed and the impact that this error has within the mitigation effort is accounted for through quantification of this error and the implementation of a strategy for mitigating the likelihood that this error causes erroneous evaluations of the vehicle's condition. Finally a full demonstration of the MERLIN method is presented within a simulation framework which includes the simulation of vehicle dynamics, pilot behavior, LOC envelope definition and real-time monitoring, and the communication of simplified recovery actions.Ph.D