Publication: Esnek Üstyapıların Mekanik Özelliklerinin Yapay Sinir Ağları Kullanılarak Geri-hesaplanması
Loading...
Files
Date
Authors
Advisor
Department
Ulaştırma Mühendisliği
Transportation Engineering
Transportation Engineering
Journal Title
Journal ISSN
Volume Title
Publisher
Fen Bilimleri Enstitüsü
Institute of Science and Technology
Institute of Science and Technology
Type
Abstract
Bu çalışmada, esnek üstyapıların mekanik özelliklerinin geri-hesaplanması, yapay sinir ağları yöntemi kullanılarak yapıldı. Ayrıca, bu modele en uygun öğrenme algoritması ve ağ mimarileri de belirlendi. Bu amaçla, 350 farklı yapay sinir ağı tasarlandı ve her biri ayrı ayrı eğitilerek test edildi. Sonuçlar gösterdi ki, Lavenberg-Marquardt and ölçekli eşleştirmeli bileşen öğrenme yöntemleri en başarılı olanlardı ve klasik öğrenme yöntemlerinden çok daha hassas sonuçlar verdiler. Ek olarak, elde edilen bulgular farklı bir üstyapı modeli üzerinde denendi ve sonuçlar önce bulunan neticeler ile uyum sağladı. Ayrıca, detaylı bir hata analizi yapılarak, eğitimi erken sonlandırmanın model başarısı üzerindeki etkisi de çalışma kapsamında incelendi. Yapay sinir ağı modeline ait işlem hızı gerçek-zamanlı geri-hesaplama yapılmasına olanak sağlamakta ve bu da kaplama peformans değerlendirmelerinde ciddi bir ekonomik ve zaman tasarrufu sağlamaktadır. Bununla birlikte, geri-hesaplama için kullanılacak yapay sinir ağı modelinde, öğrenme algoritmasının ve ağ mimarisinin seçimi ve doğru bir hata analizi yapılması büyük önem taşımaktadır.
In this study, neural network approach was applied on the backcalculation of mechanical properties of flexible pavements. In addition, the most appropriate learning technique and network architecture for this problem was also determined. For the evaluation of learning algorithm, network architecture, and network parameters, 330 different neural networks were trained and tested. Results indicated that Lavenberg-Marquardt and Scaled Conjugate Gradient algorithm are the most successful of all. In addition, the findings tested on the different pavement models, and the verification was quite successful. Besides, a comprehensive error analysis and early stopping criterion was conducted for the considered problem. The backcalculation speed of neural networks enables real-time backcalculation opportunity. Hence, serious time and money that were spent for the pavement performance evaluation processes can be saved. Nevertheless, the learning algorithm, network structure, and error analysis are crucial concepts for neural network based backcalculation.
In this study, neural network approach was applied on the backcalculation of mechanical properties of flexible pavements. In addition, the most appropriate learning technique and network architecture for this problem was also determined. For the evaluation of learning algorithm, network architecture, and network parameters, 330 different neural networks were trained and tested. Results indicated that Lavenberg-Marquardt and Scaled Conjugate Gradient algorithm are the most successful of all. In addition, the findings tested on the different pavement models, and the verification was quite successful. Besides, a comprehensive error analysis and early stopping criterion was conducted for the considered problem. The backcalculation speed of neural networks enables real-time backcalculation opportunity. Hence, serious time and money that were spent for the pavement performance evaluation processes can be saved. Nevertheless, the learning algorithm, network structure, and error analysis are crucial concepts for neural network based backcalculation.
Description
Tez (Doktora) -- İstanbul Teknik Üniversitesi, Fen Bilimleri Enstitüsü, 2004
Thesis (PhD) -- İstanbul Technical University, Institute of Science and Technology, 2004
Thesis (PhD) -- İstanbul Technical University, Institute of Science and Technology, 2004
Journal or Series
ISSN
ISBN
Rights
İTÜ tezleri telif hakkı ile korunmaktadır. Bunlar, bu kaynak üzerinden herhangi bir amaçla görüntülenebilir, ancak herhangi bir biçimde yeniden oluşturulması veya dağıtılması yazılı izin alınmadan yasaklanmıştır.
İTÜ theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.
İTÜ theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.
Keywords
Esnek üstyapı, Yapay sinir ağları, geri hesaplama, Flexible pavement, Neural networks, backcalculation
Citation
Endorsement
Review
Supplemented By
Referenced By
12
Görüntülenme
30
İndirme
Google Scholar
Scholar'da Ara ↗ Bu yayında DOI yok — Altmetric/Dimensions/PlumX/BIP! rozetleri DOI gerektirir.