Genetik algoritma kullanarak uçak park yeri optimizasyonu

The growth of aviation industry created new problems. One of these problems is the GAP (Gate Assignment Problem). Increasing aircraft traffic at large airports requires the effective use of gates. Airplane parking positions have affected the operational costs as well as the ability of transit passengers and their baggage to reach the next flight. The goal of GAP is to park aircrafts at the appropriate locations on the ground to ensure that airport parking area is used in a most efficient way. In literature, first the total walking distance of transit passengers, then the distance from the parking position to the baggage area and the total distance to the next flights of transit passengers are used as the target function. In 1980s, Babic et al. [2] did some research on this problem by using integer-programming and it got investigated by many others. In the following years, statistical models were preferred instead of deterministic models. Various simulation models were proposed by Yu Cheng [3]. Bolat [4] took operational costs as a target function instead of transit passengers or baggage distance. For the first time in 1998, GA [5] was used for GAP. The calculation of this complicated problem (GAP) is shortened using GA. In 2001, multiple targeted problems have been proposed [6]. Linkage and uniform gene exchange methods were used for GAP by Paolo [7]. In this thesis, GAP is solved by GA method. Total transit passenger walking distance is used as an objective function. This method aims to minimize airline cost and maximize customer satisfaction. This will also minimize the time delay and the number of delayed planes. The GA has been developed properly for GAP without any use of GA program package, and it has been introduced in the interface of a web page 'http://genetik1.netf13.com/home'. This paper aims the optimization of planes allocation in order to minimize the operation cost, especially in hub airports. Gate assignment problem (GAP) is a process which affects hub airports scheduling if it is not handling with a great attention. Even there are similar studies, this paper's application runs on an interactive and open to public web page. In this thesis, it is emphasized that GA gets inspiration from natural selection. These type of algorithms are successful in problem-solving because they use the nature's rules of survival. A genetic algorithm is an evolutionary algorithm. So there must be appropriate parameters of the algorithm to work properly. These parameters are chosen in accordance with GAP and the size of the problem. While coding the program, a modular structure was constructed. This program can be developed easily because changes in a specific class are only effective within its class. As a result of this, changes do not affect the general working protocol. Enhanced Permutation Code is appropriate for this study because normal permutation code is insufficient for null values in representation. OX and Partially Mapped CO (PMX) are appropriate for permutation representation but OX is better for permutation with null values. xix In this application, PHP is used for the algorithm, HTML is used for GUI and MySQL4 is used for the database. As this program is a GUI, it allows other people to go to the web-site and interact. While coding the programme, a modular structure was constructed. The user can choose any parameter and start the program. However, by using their preferred parameters, they can get top 5 results, analyze and examine them. 0data_set_gate0 table is a square matrix that indicates the distance between the gates. 0data_set_plane0 table is a square matrix that indicates the number of transit passengers on a plane. Furthermore, 'open' table is for the easy use of the program with memorizing and accessing usable parameter sets. In the end, parameters and results are kept in the "stat" table. On the background, communication with the database, the start of the algorithm and parameter checks and setup is provided from PHP. This object oriented PHP system is coded by Burak Durukan and algorithm is designed by Burak Güler. The program and the code in the thesis is stored in a CD. Running parameters and results can be seen 3.1 and 3.2 sections. During the trials, when the final step is used to get a result, the outcome turns out to be random. The algorithm does not give palpable results in the final step. The algorithm should work like an artificial intelligent so stagnant step is preferable parameter for it. Using stagnant step gives a chance for algorithm to find a solution before the final step which is determined earlier but in big numbers. The stagnant step forces algorithm to stop after getting the best result. This study shows that the best stagnant step is between 1/5 and 1/9 of the final step. Another usable parameter that affects run time of the algorithm is mutation rate. Studies show that the best mutation rate for this algorithm is between 0,06 and 0,11. Due to direct changes in solutions, higher mutation rate is unnecessary and negative for the result. Another parameter behind the best result is elitism. There have been similar studies about GAP before. However, what makes this application unique is that it can be run on a web-page and is open to public. This program is open to further development as changes in a class are only effective within its class. Changes do not affect the general working protocol.