Identification of the Actual Mission Profiles and Their Impact on the Integrated Aircraft and Airline Network Optimization
View Video Presentation: https://doi.org/10.2514/6.2021-2453.vid
The traditional aircraft design techniques implemented by the industry and academy focus on the evaluation of several disciplines envisioning the optimization of a given objective function. During this process, information about the actual missions that the designed aircraft will perform under a constrained Air Traffic Management (ATM) environment is typically not considered at a realistic level, which can lead to sub-optimal solutions. In this paper, we evaluate the impacts of incorporating the actual mission profile into the integrated aircraft and airline network simultaneous optimization process. For this, we propose a data-driven approach for mission identification and an aircraft evaluation framework that embeds a linear programming model for the integrated airline fleet and network optimization under ATM constraints. The actual mission profile is learned using data from Automatic Dependent Surveillance-Broadcast (ADS-B). We use six months of aircraft tracking data from FlightRadar24 for flights between ten of the major airports of the European Union air transportation network. An unsupervised machine learning algorithm is applied to perform a two-step clustering of flight tracks at both horizontal and vertical dimensions to identify the major routes for a given Origin-Destination pair. A flight phase identification algorithm based on fuzzy logic is used to enhance the clustering process and to obtain the final aircraft missions to be considered in the aircraft evaluation framework. In our framework, the airline network allows for direct routing (i.e., no hub is forced), and the aircraft design module is composed of several disciplines (e.g. Aerodynamics, Performance, etc) adapted to the Common Parametric Aircraft Configuration Scheme (CPACS), used as a data exchange format to enhance the interaction between them. A heuristic-based genetic algorithm is used to solve the system-of-systems optimization problem. The results show that more conservative estimates in terms of profit and direct operational cost are achieved when accounting for realistic mission profiles.