Geri Dön

An efficient design optimization framework for rf and optical applications

Başlık çevirisi mevcut değil.

  1. Tez No: 665540
  2. Yazar: ORKUN KARABAŞOĞLU
  3. Danışmanlar: DOÇ. DR. GULLU KIZILTAS
  4. Tez Türü: Yüksek Lisans
  5. Konular: Mekatronik Mühendisliği, Mechatronics Engineering
  6. Anahtar Kelimeler: Belirtilmemiş.
  7. Yıl: 2008
  8. Dil: İngilizce
  9. Üniversite: Sabancı Üniversitesi
  10. Enstitü: Fen Bilimleri Enstitüsü
  11. Ana Bilim Dalı: Mühendislik Bilimleri Ana Bilim Dalı
  12. Bilim Dalı: Belirtilmemiş.
  13. Sayfa Sayısı: 78

Özet

Metamaterials have gained considerable interest in the RF and optics community due to their unusual properties not available in nature. However, despite their proven potential in theory and some practical realizations, metamaterials are restricted to some known geometry or physical compositions such as the SRR structure. This is largely due to the natural limit imposed by intuitive design efforts and or practical realization challenges. A formal efficient framework allowing for the design of non-intuitive structures does not exist. Similar to metamaterials, artificial composite designs in literature prompt for the possibility of unique material combinations possibly in three dimensions leading to desired electromagnetic behavior such as non-reciprocality. Formal design optimization can explore unknown design degrees of freedom. However, large scale design optimization problems such as volumetric material explorations are computationally very expensive and demand high resources. This drawback can be surpassed by introducing surrogate modeling techniques into the platform as demonstrated in literature. To address this issue, in this thesis, we present an efficient design optimization framework for electromagnetic applications. The framework is based on integrating design optimization techniques with various surrogate modeling tools. The goal is to identify the device structure, both material and conductor in three dimensions, in an automated and efficient manner subject to some performance and size constraints. For the synthesis module, gradient-based optimizers such as Sequential Quadratic Programming (SQP) and global optimizers such as Genetic Algorithms are both utilized. As the analysis module, full wave electromagnetic wave analyzers, such as the HFSS, and a hybrid FE-BI based electromagnetic solver (FSDA) are integrated to the optimizers. The design framework is primarily based on interfacing the analysis tools with various surrogate based models and linking it to the chosen optimization tool. The proposed v framework is modular, hence allows for various combinations of synthesis and analysis modules within different surrogate based models. To allow for considerable speed-ups of the automated design process proposed here, automated and adaptive Design of Experiment (DOE) scheme is employed and various surrogate models are compared within the design framework with respect to their performance. Multiple surrogate modeling techniques are investigated such as Kriging, Polynomial, RBF (Radial Basis Functions), Artificial Neural Networks (ANN), Support Vector Machines (SVM), etc. Results suggest that the ANN surrogate model based design framework allows for large number of design variables and an effective exploration of the global large scale design space while RBF, SVM and Kriging are also successful at capturing resonance behavior. Multi-objective optimization method called NormalBoundary Intersection (NBI) is integrated our framework. The resulting framework is suitable for the design of volumetric material compositions of applications such as an ultra-wide bandwidth SATCOM antenna and plasmonic nano-antennas. Future work comprises the determination of the artificial material structure following a two step procedure: The effective medium of the antenna is to be determined using the proposed surrogate based design optimization framework. This should allow for extending the capabilities of the proposed framework to the design of the microstructure of metamaterials utilizing inverse topology optimization. The freedom to explore all possible design degrees of freedom and the possibility to design for the material itself are expected to open up entirely new breakthroughs in microwave and optical applications.

Özet (Çeviri)

Metamaterials have gained considerable interest in the RF and optics community due to their unusual properties not available in nature. However, despite their proven potential in theory and some practical realizations, metamaterials are restricted to some known geometry or physical compositions such as the SRR structure. This is largely due to the natural limit imposed by intuitive design efforts and or practical realization challenges. A formal efficient framework allowing for the design of non-intuitive structures does not exist. Similar to metamaterials, artificial composite designs in literature prompt for the possibility of unique material combinations possibly in three dimensions leading to desired electromagnetic behavior such as non-reciprocality. Formal design optimization can explore unknown design degrees of freedom. However, large scale design optimization problems such as volumetric material explorations are computationally very expensive and demand high resources. This drawback can be surpassed by introducing surrogate modeling techniques into the platform as demonstrated in literature. To address this issue, in this thesis, we present an efficient design optimization framework for electromagnetic applications. The framework is based on integrating design optimization techniques with various surrogate modeling tools. The goal is to identify the device structure, both material and conductor in three dimensions, in an automated and efficient manner subject to some performance and size constraints. For the synthesis module, gradient-based optimizers such as Sequential Quadratic Programming (SQP) and global optimizers such as Genetic Algorithms are both utilized. As the analysis module, full wave electromagnetic wave analyzers, such as the HFSS, and a hybrid FE-BI based electromagnetic solver (FSDA) are integrated to the optimizers. The design framework is primarily based on interfacing the analysis tools with various surrogate based models and linking it to the chosen optimization tool. The proposed v framework is modular, hence allows for various combinations of synthesis and analysis modules within different surrogate based models. To allow for considerable speed-ups of the automated design process proposed here, automated and adaptive Design of Experiment (DOE) scheme is employed and various surrogate models are compared within the design framework with respect to their performance. Multiple surrogate modeling techniques are investigated such as Kriging, Polynomial, RBF (Radial Basis Functions), Artificial Neural Networks (ANN), Support Vector Machines (SVM), etc. Results suggest that the ANN surrogate model based design framework allows for large number of design variables and an effective exploration of the global large scale design space while RBF, SVM and Kriging are also successful at capturing resonance behavior. Multi-objective optimization method called NormalBoundary Intersection (NBI) is integrated our framework. The resulting framework is suitable for the design of volumetric material compositions of applications such as an ultra-wide bandwidth SATCOM antenna and plasmonic nano-antennas. Future work comprises the determination of the artificial material structure following a two step procedure: The effective medium of the antenna is to be determined using the proposed surrogate based design optimization framework. This should allow for extending the capabilities of the proposed framework to the design of the microstructure of metamaterials utilizing inverse topology optimization. The freedom to explore all possible design degrees of freedom and the possibility to design for the material itself are expected to open up entirely new breakthroughs in microwave and optical applications.

Benzer Tezler

  1. Self-organized network management model for next generation wireless heterogeneous systems

    Yeni nesil kablosuz çoktürel sistemlerde kendini düzenleyen ağ yönetim modeli

    ÖZGÜR UMUT AKGÜL

    Yüksek Lisans

    İngilizce

    İngilizce

    2014

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve Kontrolİstanbul Teknik Üniversitesi

    Bilgisayar Mühendisliği Ana Bilim Dalı

    YRD. DOÇ. DR. BERK CANBERK

  2. Visible light positioning systems: Fundamental limits, algorithms and resource allocation approaches

    Görünür ışık konumlandırma sistemleri: Temel sınırlar, algoritmalar ve kaynak tahsisi yaklaşımları

    MUSA FURKAN KESKİN

    Doktora

    İngilizce

    İngilizce

    2018

    Elektrik ve Elektronik Mühendisliğiİhsan Doğramacı Bilkent Üniversitesi

    Elektrik-Elektronik Mühendisliği Ana Bilim Dalı

    PROF. DR. SİNAN GEZİCİ

  3. Full duplex hybrid acoustic/RF communication for underwater networked control systems

    Su altı ağ bağlantılı kontrol sistemleri için tam çift yönlü hibrit akustik/RF haberleşme

    SAEED NOURIZADEH AZAR

    Doktora

    İngilizce

    İngilizce

    2022

    Bilgisayar Mühendisliği Bilimleri-Bilgisayar ve KontrolSabancı Üniversitesi

    Bilgisayar Mühendisliği Ana Bilim Dalı

    PROF. DR. ÖZGÜR GÜRBÜZ ÜNLÜYURT

    DOÇ. DR. AHMET ONAT

  4. Sistemli mimari tasarım sürecinde üretken bir yapay yardımcı önerisi

    Proposal for a productive artificial aid in the systematic architectural design process

    BETÜL ŞEN

    Yüksek Lisans

    Türkçe

    Türkçe

    2024

    Mimarlıkİstanbul Teknik Üniversitesi

    Bilişim Ana Bilim Dalı

    PROF. DR. YÜKSEL DEMİR

  5. Relaying opportunities for wireless networks by applying network coding

    Kablosuz ağlar için ağ kodlamalı aktarma fırsatları

    SEMİHA TEDİK BAŞARAN

    Doktora

    İngilizce

    İngilizce

    2019

    Elektrik ve Elektronik Mühendisliğiİstanbul Teknik Üniversitesi

    Elektronik ve Haberleşme Mühendisliği Ana Bilim Dalı

    PROF. DR. GÜNEŞ ZEYNEP KARABULUT KURT