Statistically Modeling Fuel Consumption with Heteroscedastic Data
Aircraft operate in unpredictable environmental conditions. As a result, autopilot design is difficult, as optimal responses cannot be anticipated for all conditions. Consequently, the autopilot might overcorrect for conditions, using more fuel than necessary. By analyzing performance data on a subject aircraft, the relationships between environmental condition variables and fuel consumption using linear regression models have been characterized. These relationships are accurate, even though the data is non-normal and heteroscedastic.