Geri Dön

Harrıs Hawks optimizasyon algoritması ile sürü davranışlarının modellenmesi ve analizi

Modeling and analysis of swarm behavior with Harris Hawks optimization algorithm

  1. Tez No: 812943
  2. Yazar: GÖZDE SOFUOĞLU
  3. Danışmanlar: DR. ÖĞR. ÜYESİ GÖKHAN ATALI
  4. Tez Türü: Yüksek Lisans
  5. Konular: Mekatronik Mühendisliği, Mechatronics Engineering
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2023
  8. Dil: Türkçe
  9. Üniversite: Sakarya Uygulamalı Bilimler Üniversitesi
  10. Enstitü: Lisansüstü Eğitim Enstitüsü
  11. Ana Bilim Dalı: Mekatronik Mühendisliği Ana Bilim Dalı
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 101

Özet

Bu tez çalışmasında, metasezgisel optimizasyon yöntemlerinden olan ve doğada en zeki yırtıcılar olarak bilinen harris şahinlerinin avlanma davranışından esinlenerek harris hawks optimizasyon (HHO) algoritması üzerinde iyileştirmeler yapılmış ve benzetimli tavlama algoritması (SA) ile karşılaştırması yapılarak sürü davranışları incelenmiştir. Çalışmada ilk olarak belli sınırlar içerisinde rastgele ortaya çıkan ajanların arama ve yok etme davranışının optimize edilmesi amaçlanmıştır. Bu amaç doğrultusunda model MATLAB ortamında oluşturulmuştur. Oluşturulan modelde ajan ve sürü tanımlamaları yapılmış ve her birine rastgele ağırlık katsayıları eklenerek sürünün doğru orantılı şekilde önce düşük katsayılı ajan için ardından yüksek katsayılı ajan için arama ve yok etme davranışı incelenmiştir. Bu kapsamda klasik HHO erken yakınsama ve arama davranışındaki eksiklikler sebebiyle HHO yapısı korunarak, kaotik lojistik harita fonksiyonu ile sürü davranış incelemesinin yapılabilmesini sağlayan arşiv yapısı eklenerek iyileştirilmiş kaotik harris hawks optimizasyon algoritması (IC-HHO) önerilmiştir. IC-HHO algoritmasının performans ölçümlemesi, model sonucu ilk oluşturulan rassal çözümle ve SA algoritması ile karşılaştırılmıştır. Tez çalışmasında 3 farklı deneysel çalışma gerçekleştirilmiştir. Her bir deneysel çalışmada model içerisinden girdi verileri değiştirilerek yeni kurgu yapıları oluşturulmuştur. Ardından IC-HHO ile SA algoritmaları her bir deneysel çalışmada 30 kez çalıştırılarak sonuçlar kayıt altına alınmıştır. Deneysel çalışmalar sonucunda algoritmaların arama davranışında sergilediği sürü davranış incelemesi ve optimum mesafe tespit sonuçları hem grafiksel hemde istatistiksel olarak gösterilmiştir. Deneysel sonuçlar incelendiğinde tüm deneysel çalışma sonuçlarında IC-HHO algoritması SA algoritmasına kıyasla daha üstün performans gösterdiği görülmüştür. Son olarak sürünün ve ajanların her iterasyon sonucu konum bilgisi kayıt altına alınarak davranış modellemesi incelenmiştir. IC-HHO algoritması SA algoritmasına göre davranışsal olarak geniş çözüm aralığına ve arama davranışındaki üstün performansından dolayı daha başarılı sonuçlar gösterdiği gözlemlenmiştir. Böylece IC-HHO'nun başarısını artırmaya yönelik farklı optimizasyon problemlerine uyarlanmasında varyantları geliştirilebilir ve performans karşılaştırmasında diğer optimize edicilerle kıyaslanarak yüksek çözüm üreten problemler için uyarlanabilir olduğu saptanmıştır.

Özet (Çeviri)

The subject of herd intelligence has become widespread with the examination of living groups living in herds in nature. The swarm intelligence approach is an approach that allows research studies in the field of engineering, such as examining and analyzing multiple behaviors. In this study, it is aimed to examine the swarm behaviors against unwanted agents that may occur within certain limits on swarm robots. In this direction, it is aimed to work on metaheuristic algorithms accepted in the literature. In the study, which is based on the destruction of randomly occurring unwanted agents within certain limits, the herd's ability to scan the environment in a short time and eliminate the danger will be examined. handled with different optimization algorithms in order to optimally destroy unwanted agents in the environment that occurs at different locations and at different weight coefficients. Optimum herd behavior will be determined by comparing the results of the experimental approach, which will be performed at different possibilities each time. In this direction, swarm behavior studies in different experimental studies were carried out using the harris hawk optimization (HHO) algorithm and the simulated annealing (SA) algorithm. The HHO algorithm, which is one of the meta-heuristic optimization methods developed by inspired by the hunting behavior of harris hawks, which is one of the most intelligent predators in nature high performance solution quality and thanks to its flexible structurehas been the of many researchers. In this thesis, an improved chaotic harris hawks optimization (IC-HHO) algorithm, inspired by the HHO algorithm, is proposed and the swarm behavior is examined by comparing it with the SA algorithm. In the study, it was first aimed to optimize the search and destruction behavior of randomly occurring agents within certain limits. For this purpose, model and other studies were created in MATLAB environment. In the created model, agent and swarm definitions were made and random weight coefficients were added to each of them, and the search and destruction behavior of the swarm, first for the low coefficient agent and then for the high coefficient agent was investigated. The model works in a two-dimensional plane. As initial data, xmin, xmax, ymin, ymax values, together with the number of agents and herds, were entered. After the data entries in the model setup and related calculations are made, the first solution is performed randomly. After the random initial solution, swarm distribution is performed. Then, the swarm distribution is measured first for the low coefficient agent and then for the high coefficient agent. Finally, the total distance traveled is displayed. In this step of the model, since the distances between agents, between swarms and between swarms and agents are known, the total distance is calculated. Finally, as the output of the model, the swarm distribution obtained with the first random solution and its distance to the stationary agents are found. At this stage of the model, the behavior of destroying agents is not added. The use of weight coefficients was used to determine the individual distance of the swarm relative to the agent weight coefficients. In this context, in order to measure the search and extermination behavior of the swarm within the framework of the constraints we have built on the model, a chaotic-based developed IC-HHO algorithm has been proposed due to the early convergence of the classical HHO. In the proposed method, the local search capability of the HHO was increased by increasing the population diversity thanks to the chaotic array, and the early convergence of the exploration/search process was prevented. In addition, in the proposed approach, an archive structure where candidate solutions are kept in each iteration has been added in order to be able to adapt the model and be used in the analysis processes. In this way, the imbalance in the local search and exploitation stages in the classical HHO was tried to be eliminated, and the herd behavior analysis was provided thanks to the added archive structure. The IC-HHO algorithm, which was developed by preserving the structure of the HHO algorithm, has been brought to the literature Finally, since solutions are produced for single-purpose problems by adhering to the classical HHO structure, the behavior of the swarm for the two agents in the model is separated according to the weight coefficients of the agents. Searching and destroying behaviors were observed and recorded, first the low-weight agent and then the high-weight agent, respectively. In the second step of the study, SA optimization, another popular algorithm in the literature, was used to examine the herd behavior. This algorithm was chosen because of the probabilistic approach of the simulated annealing algorithm, its easy adaptability to problems in a controlled structure, and its adaptability to combinatorial optimization problems by combining more than one solution. In addition, since it can make random changes in the analysis processes, reducing the probability of falling into the local optimum, the random change mechanism between solutions is provided with the swap function in order to improve the answers in the SA algorithm. In the study, the data coming from the model were defined and the search behavior was performed first for the low coefficient agent and then for the high coefficient agent in order to adapt it to the single-objective problem layout in the HHO algorithm structure. The search behavior performed separately for these two agents has been tried to be optimized without changing the SA algorithm structure. Since the agents are considered immobile in the SA algorithm structure, the behavior of destroying the agents with the power change as a result of each iteration is not applied. Performance measurement of the IC-HHO algorithm, the model result was compared with the first generated random solution and the SA algorithm. In this direction, 3 different experimental studies were carried out. In each experimental study, new editing structures were created by changing the input data from the model. Then, the IC-HHO and SA algorithms were run 30 times in each experimental study, and the results were recorded. As a result of experimental studies, swarm behavior analysis and optimum distance detection results of algorithms in search behavior are shown both graphically and statistically. As a result of the first experimental study, the total distance obtained from the first solution of the model was found to be 2.0735E+01. When the standard deviation data of the Experiment 1 results were evaluated, it was found as 1.3577E+00 in the IC-HHO algorithm and 2.0828E+00 in the SA algorithm. For this reason, it has been seen that the IC-HHO algorithm gives more successful results compared to the SA algorithm. As a result of the second experimental study, the total distance obtained from the first solution of the model was 3.4073E+01. When the standard deviation data of Experiment 2 were evaluated, it was found as 1.5682E+00 in the IC-HHO algorithm and 4.9603E+00 in the SA algorithm. For this reason, it has been seen that the IC-HHO algorithm gives more successful results compared to the SA algorithm. As a result of the third experimental study, the total distance obtained from the first solution of the model was 5.6356E+01. When the standard deviation data of the Experiment 3 results were evaluated, it was found as 4.3872E+00 in the IC-HHO algorithm and 5.2409E+00 in the SA algorithm. For this reason, it has been seen that the IC-HHO algorithm gives more successful results compared to the SA algorithm. As a result, when statistical data were evaluated, it was observed that the proposed IC-HHO algorithm gave more successful results compared to the SA algorithm. Thus, it has been determined that the IC-HHO algorithm can be used in complex engineering problems and in producing high-level solutions when the solution space is expanded and improvements are made in the search behavior. In addition, thanks to its easily adaptable structure, performance evaluation can be made by comparing the IC-HHO algorithm with other popular metaheuristics and test functions in the literature, and different variant studies can be developed to increase the success rate.

Benzer Tezler

  1. Yapısal optimizasyon problemlerinin çözümü için yeni bir hibrid optimizasyon yönteminin geliştirilmesi

    Development of a new hybrid optimization method for solution of structural optimization problems

    ERHAN DÜZGÜN

    Doktora

    Türkçe

    Türkçe

    2024

    Makine MühendisliğiTOBB Ekonomi ve Teknoloji Üniversitesi

    Makine Mühendisliği Ana Bilim Dalı

    PROF. DR. ERDEM ACAR

  2. Sezgisel PID kontrolör tasarımı ve DC motor simülasyonu

    Metaheuristic PID controller design and simulation of DC motor

    ABDULMUTALİP YILDIRIM

    Yüksek Lisans

    Türkçe

    Türkçe

    2021

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolFırat Üniversitesi

    Yazılım Mühendisliği Ana Bilim Dalı

    PROF. DR. SONER ÖZGEN

    PROF. DR. ERKAN TANYILDIZI

  3. Real time calculation of shem angles by metaheuristic algorithms for variable frequency motor drives

    Değişken frekanslı motor sürücüleri için metasezgisel algoritmalar kullanılarak shem açılarının gerçek zamanlı hesaplanması

    GAMZE NALÇACI

    Doktora

    İngilizce

    İngilizce

    2021

    Elektrik ve Elektronik MühendisliğiOrta Doğu Teknik Üniversitesi

    Elektrik ve Elektronik Mühendisliği Ana Bilim Dalı

    PROF. DR. MUAMMER ERMİŞ

    PROF. DR. IŞIK ÇADIRCI

  4. Meta-sezgisel algoritmalar ile Trafik Sinyal Optimizasyonu

    Traffic Signal Optimization with meta-heuristic algorithms

    OLCAY ADIYAMAN

    Yüksek Lisans

    Türkçe

    Türkçe

    2023

    Elektrik ve Elektronik MühendisliğiKayseri Üniversitesi

    Elektrik-Elektronik Mühendisliği Ana Bilim Dalı

    DOÇ. DR. ALİ DURMUŞ

  5. Metasezgisel optimizasyon algoritmaları kullanarak çok rotorlu insansız hava aracı ile faydalı yük al bırak görevi için rota planlama

    Path planning for payload pick-drop mission with multi-rotor unmanned aerial vehicle using metaheuristic optimization algorithms

    EGEMEN BELGE

    Doktora

    Türkçe

    Türkçe

    2024

    Elektrik ve Elektronik MühendisliğiZonguldak Bülent Ecevit Üniversitesi

    Elektrik-Elektronik Mühendisliği Ana Bilim Dalı

    PROF. DR. RIFAT HACIOĞLU