Evolucijska optimizacija geometrije profila letalskega krila

BELTRAM, SIMON (author), Robnik Šikonja, Marko (mentor), Papa, Gregor (comentor)

V diplomski nalogi je obravnavana hevristična optimizacija geometrije pro- fila krila prostoletečih modelov letal kategorije F1A. Preveriti želimo ali je mogoče s pomočjo hevristične optimizacije generirati profil, ki bi bil boljši od tistih, ki so trenutno v uporabi. V ta namen je bila razvita programska rešitev, ki ima implementirane optimizacijske algoritme diferencialna evolu- cija (DE), metoda roja delcev (PSO) ter metoda ognjemeta (FA). Profili, ki so trenutno v uporabi, se imenujejo profili LDA (low-drag airfoil) in so bili razviti na podlagi znanja ter izkušenj. Profili LDA lahko dosegajo visoko hitrost v fazi pridobivanja višine, a nekoliko slabše prosto letijo v primerjavi s predhodniki. S pomočjo hevrističnih optimizacijskih metod poskušamo pri- dobiti tak profil, ki bi lahko obe nalogi opravljal dovolj dobro. Vsi trije algoritmi so bili implementirani v programskem jeziku java, s pomočjo programske knjižnice JMetal, ki jo je bilo potrebno ustrezno prilagoditi. Največja sprememba je način predstavitve osebkov. V našem primeru je pro- fil letalskega krila predstavljen s parom Bézierjevih krivulj, ena za zgornjo in druga za spodnjo polovico krila. Tak način predstavitve nam da veliko kon- trole in svobode za ceno morebitnih nepravilnih oblik (zaradi naključnosti). Zato je potrebno pri generiranju profilov paziti tudi na konstrukcijske omeji- tve (pretanki ali predebeli profili, itd.). Profile ocenjujemo s pomočjo Xfoil [3] programa, ki preračuna karakteristike profila. Izkazalo se je, da upoštevanje le enega kriterija kakovosti ni dovolj, zato bi bilo smiselno upoštevati še druge parametre in implementirati večkriterijski različici algoritmov.

• Language: Slovenian
• Keywords: hevristična optimizacija, diferencialna evolucija, optimi- zacija z metodo ognjemeta, optimizacija z rojem delcev, profil letalskega krila.
• Work type: Bachelor thesis/paper
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2017
• PID: 20.500.12556/RUL-91344
• Publication date in RUL: 28.03.2017

Secondary language

• Language: English
• Title: Evolutionary optimization of aerofoil profile

Abstract

This thesis deals with the heuristic optimization of an F1A free flight airfoil. The goal is to generate an airfoil that is better than the ones currently in use. We implemented three optimization algorithms, differential evolution (DE), Firework algorithm (FA) and Particle swarm optimization (PSO). Currently used airfoils, called LDA (Low-drag airfoil) are able to obtain high speed in the phase of climbing but in the free flight phase they fly poorly compared to their predecessors. With the help of optimization methods we aim to find an airfoil that performs both tasks well. All heuristic algorithms were developed in java programming language us- ing a customized JMetal library. Each airfoil is presented with two Bézier curves, one for the upper and one for the lower half of the foil. This kind of presentation gives us more freedom and control, but it comes at a price - the software may generate irregular shapes due to randomness. That is why it is important to pay attention to structural constraints (i.e. airfoil thickness). Airfoils are evaluated using Xfoil [3] which calculates the characteristics of an airfoil. Our experiments show that L/D ratio is not a good indicator of airfoil quality and one has to take into account other parameters as well. A multi-objective criteria function would therefore be advisable.

• Keywords: heuristic optimization, differential evolution, firework algorithm, particle swarm optimization, airfoil.