Köprü Ayaklarında Meydana Gelen Yerel Oyulmaların Veri Analiz Yöntemleri Kullanılarak İncelenmesi

Abstract

Köprüler, ulaşım sürekliliğinin sağlanması için gerekli yapılardır. Özellikle akarsuları geçen köprülerin alternatif yöntemlere göre çok daha işlevsel olmaları bu yapıların önemini bir kat daha artırmaktadır. Günlük ulaşımda bu denli önemli olan köprülerin tasarımında genel olarak yapısal faktörler gözönünde bulundurulmuş, hidrolik etkenlere ise gereken önem verilmemiştir. Ancak durum ayrıntılı olarak incelendiğinde yapısal etkilerin yanısıra rüzgar, deprem, taşkın vb. birçok etkenin köprü stabilitesine zarar verdiği görülmüştür. Özellikle mertebesi kolayca kestirilemeyen ve meydana gelme mekanizması bir hayli karmaşık olan hidrolik etkenler ihmal edildiğinde ortaya geri dönüşü bulunmayan olumsuzluklar çıkmaktadır. Söz konusu hidrolik etkenlerin başında, çözülmesi zor bir yapıya sahip oyulma kavramı göze çarpmaktadır. Son zamanlarda yapılan çalışmalar köprü yıkılmalarında oyulma faktörünün çok önemli olduğunu göstermiştir. Öyle ki son 30 yılda Amerika Birleşik Devletleri’nde 1000’den fazla köprü yıkılmış ve bunların yaklaşık %60’ının oyulma sebepli olduğu araştırmacılar tarafından saptanmıştır. Oyulma probleminin karmaşık bir süreç olması, yapılan araştırmaların önemini daha da artırmıştır. Bu araştırmalar oyulmanın hangi faktörlere bağlı olarak değiştiğini genel olarak açıklamaktadır. Köprü ayaklarında meydana gelen oyulmaların asli sebepleri arasında köprü ayak çapı ve şekli, akıntı hızı, yatak malzemesi çeşidi gösterilebilir. Süreç bu sebeplere bağlı olarak teker teker incelendiğinde daha net sonuçlara ulaşılabilir. Aslında oyulma mekanizması birçok etkene aynı anda bağlı olarak gerçekleşmektedir. Bu etkenlerin arasındaki bağımlılık derecesini bulmak çok zor olduğundan dolayı bunları bağımsız düşünüp ayrı ayrı incelemek ve daha sonra üst üste çakıştırılmış (süperpoze) hale getirmek yapılabilecek en sağlıklı analiz olarak görülmektedir. Gerçeğe en yakın modeli kurmak için, köprü ayak oyulması etkenlerinin çeşitli modeller kurularak incelenmesi ve hassasiyet analizi yapılması gerekmektedir. Bu çalışmada köprü ayaklarında meydana gelen oyulmalara ait bazı araştırmacılar tarafından yapılan deneylere ait veriler alınmış, yapay sinir ağları ve bulanık mantık çıkarım sistemi kullanılarak bu veriler analiz edilmiş ve elde edilen sonuçlar literatürde sıklıkla kullanılan deneysel formüllerin performanslarıyla karşılaştırılmıştır. Veri analiz yöntemleri kullanılarak elde edilen en iyi modellerin performansı, her bir girdide yapılan hasssasiyet analizleriyle ayrıca sınanmıştır. Çalışmanın 1. bölümünde problem genel olarak tanıtılmıştır. 2. bölümde oyulma konusu ile ilgili şimdiye kadar yapılan çalışmalar, alınan önlemler ve ulaşılan sonuçlar ayrıntılı bir şekilde aktarılmıştır. 3. bölümde bulanık mantık çıkarım sistemi ve yapay sinir ağları yöntemlerinin mantık altyapıları açıklanmış ve kullanım alanları hakkında ayrıntılı bilgi verilmiştir. 4. Bölümde ise, bahsedilen veri analiz yöntemleri eldeki deneysel verileri analiz etme noktasında kullanılmıştır. Analiz sonucu elde edilen bulgular ve deneysel formüllerle elde edilen sonuçlarla karşılaştırılmıştır. 5. bölümde ise yapılan çalışma neticesinde çıkarılabilecek sonuçlar ve öneriler sunulmuştur.

Bridges are the essential parts of transportation networks. Especially river bridges are more advantageous than alternative ways. Up to the present, usually structural factors have been considered but hydraulic factors haven’t been. However, while the condition analyzed in detail, there are many factors that harm the stability of bridges like wind, earthquake, flood etc. as well as structural ones. In particular hydraulic factors have a very complex mechanism that can not be solved easily. Hence, the scour mechanism has to be cared for preventing bad results. It is a common failure of bridge that scouring at bridge piers, abutments or supports. The scour concept has a very complicated mechanism and it is the most difficult hydraulic problem to solve in bridges as mentioned above. Especially the last studies show that scour is the most important factor in bridge failures. Such that there are more than 1000 bridges have collapsed in United States during last 30 years time and 60% of these caused by scour interaction according to researchers. Scour mechanism is genereally divided into two parts. First is clear water scour and the other is live bed scour. The main difference between these is flow velocity. If critical velocity exceeded, which is the peak point of clear water scour condition, live bed scour starts. When the flow velocity values lower than the critical velocity, clear water scour occurs. Local scour at bridge piers occurs when the flow velocity exceeds about half of the critical velocity. The vortexes are the another important factor in scour mechanism. Horse shoe vortex is the most effective factor in scour. The basic principle of horse shoe vortex is, the flow moves at the base of the piers when flow approaches pier. A portion of flow moves down to the pier and interacts with the river bed. This interaction between the horse shoe vortex and the down flow commences the scour. Scour phenomenon is a very complicated process. Because of this, the importance of researchs about scour has increased. Also these researchs can explain how the scour changes and depends on which factors. The main factors of bridge pier scour can be shown as diameter and shape of the pier, flow velocity and the type of bed material. When the scour process analyzed individually, conclusions can be more clearer. Actually scouring mechanism depends on many factors simultaneously. It is difficult to find the dependence between these factors, so thinking independence them first and then making a superpose is a good analyse for scour. Therefore, the best real model can be built with the most dependent factors. Analysis of data is a process of inspecting, transforming and modeling data with the goal of discovering useful information about unknown events, get some conclusions, and supporting decision making. There are many data analysis methods using many years. Fuzzy logic inference system and artificial neural network are some of the useful data analysis methods. Fuzzy logic method has used first in 1960s. The main idea is grading the membership functions of system components. When the fuzzy logic compared to classical logic, fuzzy logic variables may have different values that ranges in degree between 0 and 1. Classical logic only permits propositions having a value of true or false. It accepts only an absolute, immutable, mathematical truth. However, there exist certain propositions with variable answers, such as asking different people to identify a color or smell. Moreover, fuzzy set theory describes fuzzy operators on fuzzy sets. The problem in applying that is the appropriate fuzzy operator may not be known from user. Because of this, fuzzy logic constructs some fuzzy logic rule base using IF-THEN rules. In some disciplines, artificial neural networks are artificial intelligence models inspired by animals nervous systems (in particular the brain) that are capable of machine learning and pattern recognition. They are usually presented as systems of interconnected neurons that can compute output values from inputs by feeding information through the network. Artificial neural networks (ANNs); these are essentially simple mathematical models defining some functions may be linear or non-linear. Furthermore, models are also intimately associated with a particular learning algorithm or learning rule. A commonly used parameter is the mean-squared error, which tries to minimize the average squared error between the network s output, f(a), and the target value b over all the example pairs. When one tries to minimize this parameter using gradient descent for the class of neural networks called multilayer perceptrons (MLP), one obtains the common and well-known backpropagation algorithm for training neural networks. The aim of this study is to solve the effects of complicated scour mechanism parameters. There are many parameters defined from researchers that effect the scour occurrence. In this study, the dataset has six inputs and one output. Because of this, the effects of these inputs has investigated in this study. In this study, the experimental data about scour obtained and examined with fuzzy logic inference system and artificial neural networks. Also two empirical scour calculation formula has used for comparing the calculated, estimated and measured values. In fuzzy logic model, the data were divided into two parts so called training and testing data. The system try to learn the model via training data. The learning examined on testing data. This methodology used to find the best estimation model with many membership functions. Takagi-Sugeno model used iin this investigation. Artificial neural networks model has tried to minimize the average squared error between the network s output. Besides, the network trained with Levenberg-Marquardt backpropagation algorithm. The number of nodes in the hidden layer and network control parameters were determined through calibration. In the first section, the scour mechanism identified in general. In the second section, the studies ever made was described which are countermeasures, predictions, models and monitoring systems. Many countermasures has been tried to minimize the scour occurrence and researchers recommends the riprap layer for the best solution. In the third section, the principles of fuzzy logic inference system and artificial neural networks has described. Also detailed information has given about application areas. In the fourth part, the data analysis methods mentioned above, used for analysing the obtained data from some researchers. The outcomes of these analysis compared with the results obtained from emprical formulas. The scatter diagrams were drawn for observed and estimated results. If estimations are correct the data is stacked on the 450 line. In the last part, some results and suggestions presented. As a result of study, the most important inputs for equilibrium scour depth occurrence has determined. Pier diameter, crticical to the flow velocity ratio, the average flow depth and the average flow velocity model ise the best model for artificial neural network analysis. Similarly pier diameter, crticical to the flow velocity ratio and the average flow velocity model is the best model for fuzzy logic inference system analysis. It has also determined that the best model is the simplest model.