Abstract

large number of stakeholders exist in the modern air traffic management ecosystem. Air transportation studies benefit from collaboration and the sharing of knowledge and findings between these different players. However, not all parties have equal access to information. Due to the lack of open-source tools and models, it is not always possible to undertake comparative studies and to repeat experiments. The barriers to accessing proprietary tools and models create major limitations in the field of air traffic management research. This dissertation investigates the methods necessary to construct an aircraft performance model based on open data, which can be used freely and redistributed without restrictions. The primary data source presented in this dissertation is aircraft surveillance data that can be intercepted openly with little to no restriction in most regions of the world. The eleven chapters in this dissertation follow the sequence of open data, open models, and performance estimations. This order corresponds to the three main parts of the dissertation. In the first part of the dissertation, open surveillance data is explored. Methods are developed to decode and process this data. Extraction of information is also made possible thanks to machine learning algorithms. The second part of the dissertation examines the main components of the open aircraft performance model. Models related to kinematics, thrust, drag polar, fuel flow, and weather are investigated. The third part of the dissertation looks into the possibility of using surveillance data to estimate aircraft performance parameters, for example, aircraft turn performance, aircraft mass, and thrust settings, for individual flights. With the goal of making future air traffic management studies more transparent, comparable, and reproducible, the models and tools proposed in this dissertation are fully open. The final aircraft performance model, OpenAP, proposed in this dissertation has proven to be an efficient open alternative to current closed-source models.

Abstract

Considering the expected air traffic growth, the innovation and development of new tools able to allow a more efficient and safer way of managing aircraft operations is necessary to achieve future expectations. In this context, it is important to be capable of accurately predict aircraft trajectories to ensure efficient aircraft operations (e.g., planning and dispatching of flights, in-flight trajectory prediction, etc.) as well as to make more robust the Air Traffic Management (ATM) system (which includes ground based systems for ATC, predicting demand in ATC sectors, etc). The way of predicting them is based on aircraft performance models (APM), i.e., sets of equations that allow to model the aircraft performance according to some specific coefficients that depend on the aircraft carrying out the flight. Therefore, the predicted trajectories accuracy will depend directly on the aircraft performance model used. If the APM does not reflect the reality, the predicted trajectories will not be accurate enough. Moreover, since these trajectories will not be longer optimal according to the real performance model, the performance in cost-efficiency and the environmental impact of aircraft operations will be decreased. Then, it is required the use of aircraft performance models as realistic as possible. The objective of this Master Thesis is the design of an algorithm capable to estimate the coefficients of the functions describing the aircraft performance model considered, which will be the Base of Aircraft Data (BADA), from Quick Access Recorder (QAR) flight data. The idea is to obtain the real coefficients values of each specific tail number aircraft to accurately model their performance for trajectory prediction purposes. To perform the coefficients' estimation, real flight data has been used, consisting in a set of trajectories carried out by more than 30 different tail number aircrafts and obtained from their respective QAR. Among these aircrafts, only one (i.e, its trajectories) will be selected to design the estimation algorithm, which consist in a nonlinear least squares optimization problem. The output will be the coefficients value that better fits to the QAR data used, i.e, that minimize the difference between the real trajectories and the generated predicted trajectories as a result of applying the estimated coefficient's value to the BADA model equations. The results show a good estimation accuracy for the four aircrafts analysed, obtaining a very low prediction error in all cases, which proves the reliability of the algorithm and the necessity of tailoring the existing performance models (in this case BADA).