The work presented in this thesis is applied in the field of aircraft flight trajectory optimization, approached as a flight-planning problem. The objective of the optimization is to determine an optimal flight plan, which minimizes a selected cost function and satisfies all the imposed constraints. The optimization takes into account the particular aircraft performance data and flight configuration (load, fuel quantity, etc.), initial and final points (latitudes, longitudes, and altitudes) of the flight segment to be optimized, atmospheric conditions along the flight trajectory, as well as optimization and navigation constraints. It was assumed in this work that the lateral component of a flight plan was composed by a set of sub-segments, constructed by selecting adjacent nodes from a routing grid. The routing grid was constructed based on the orthodromic route between the initial and final points of the segment to be optimized, a selected maximum lateral deviation from the orthodromic route, and a maximum sea level distance between the grid nodes.
The first research subject concerned a new atmospheric data model that defines the variation of the atmospheric parameters as functions of time in selected points along the lateral flight trajectory or in the nodes of a routing grid, at a selected altitude. The model was constructed based on the forecast data provided by the Meteorological Agencies, in GRIB2 data format, and defined in the nodes of a 4D grid (geographic location, altitude, and time). As a result, an atmospheric parameter value in an atmospheric data definition point (geographic location and altitude), at the time instance of interest, was obtained by a one-dimensional linear interpolation. Test results showed that, compared with the classic four-dimensional linear interpolation from the GRIB forecast data, the proposed model yielded identical atmospheric parameters values (differences of the order of 10-14) and, on average, it was six times faster. Therefore, by using the proposed atmospheric data model it would be possible to perform an optimization faster or to evaluate more candidate flight plans during the allotted execution time, which would yield better optimization results. The proposed model can be extended by generating the model data for each altitude from a set of altitudes of interest.
The second investigation evaluates the performance of a new optimization method, based on genetic algorithms, where both the lateral and the vertical components of the flight plan are subjected to optimization. In this study, the routing grid for the lateral component of the candidate flight plans was constructed according to the methodology presented in the first investigation. The family of vertical flight plans was constructed according to a selected structure and topology. The results were compared with a reference flight plan, obtained as the optimal profile (speed optimization) for a flight along the flight track and altitude profile of a real flight, retrieved from the FlightAware website. Subsequently, another investigation analyzed the effects of performing flight plan corrections (altitude – speed profile) relative to the aircraft flight envelope (correction of the candidate flight plan parameters, so that the aircraft flight parameters would remain within the flight envelope limits), on the optimization results and execution time. A total of 60 tests were performed, composed of 10 test runs for each of the six cost index values considered in the evaluation. The results showed that, by performing the flight plan corrections relative to the aircraft flight envelope, the computation times increase by a factor larger than two and the results are less optimal. Relative to the reference flight plan, the proposed optimization method, in which the candidate flight plans were not corrected, yielded a total cost reduction between 1.598% and 3.97%.
The third investigation evaluates a new flight plan optimization method / approach, derived from the Non-dominated Sorting Genetic Algorithm II multi-objective optimization method. The proposed method applies to the case where a crossing time is desired / expected to be imposed at the final point of the segment under optimization (Required Time of Arrival). The time constraint value could be a preferred crossing time instance selected by the flight planner or, it could result from a negotiation with the Air Traffic Management System. The proposed method identifies, in parallel, a set of optimal flight plans corresponding to a set of selected contiguous flight time constraints (“windows”) imposed at the final point of the flight segment to be optimized. The advantage of the proposed method is that decision makers can select the flight plan that best suits their criteria and, if rejected by the Air Traffic Management system, they can select the next best flight plan from the set of solutions without having to perform a new optimization. Seven method variants were evaluated, and 10 test runs were performed for each variant. The tests considered the case where 31 contiguous time constraint windows were imposed at the final point of the segment under optimization. Test results showed a very good convergence of the solutions. For five method variants the maximum fuel burn differences relative to the “global” minimum for a time constraint value (for all the method variants and all the test runs) were less than 90 kg of fuel (0.14%). The worst optimization method found optimal flight plans that yielded fuel burns with a maximum of 321 kg (0.56%) more than the “global” optimum.
| Date | 27 Jan 2021 |
|---|
| Original language | American English |
|---|
| Awarding Institution | - École de technologie supérieure
|
|---|
| Supervisor | Ruxandra Botez (Supervisor) |
|---|