Constraint-Based Analysis of Aircraft Operations

The analysis of flight data presents us with the opportunity of sophisticated insights into the functioning of aircraft systems and into the effects of aircraft operations. However, the complexity and size of the data acquired through flight data logging represent a challenge to the most complex tools to date. The effort proposed herein sets forth a multi-phase approach for the efficient application of existing data processing, visualization and analysis techniques for the identification and characterization of safety-relevant flight conditions, and for the identification of their precursors. The approach starts by reducing the voluminous flight data into a feature-based encoding, where features used for each flight parameter type are selected to match its specific variation. The second processing phase allows domain experts to efficiently inspect the flight data starting from detected exceedances, and to formulate constraint-based hypotheses regarding co-occurring events, and possible precursors. The third phase uses the hypotheses developed during the second phase to apply constraint-based data mining techniques for the identification of patterns associated with exceedances, and to identify the precursor sets for safety-relevant flight conditions. The description of the proposed effort also includes preliminary implementation results for selected prototype components, and establishes the feasibility of the investigated approach.