Agregalarin Fiziksel Özelliklerinden Yola Çikilarak Beton Dayanimlarinin Yapay Sinir Ağlariyla Kestirilmesi

Abstract

Agreganın sahip olduğu fiziksel ve mekanik özelliklerin betonun dayanım özelliklerine etkisinin belirlenmesi ancak yapılan deneylerle mümkündür. Deneysel çalışmalar, uzun süreçlerde yapılabilen, malzeme harcanan ve ekonomik yükümlülük getiren, aynı zamanda teknik personel gerektiren çalışmalardır. Bu yüzden yapay sinir ağları bu kayıpların ve gereksinimlerin daha aza indirgenebilir olduğunu gösterilmiştir. Bu çalışmada Marmara bölgesinin farklı lokasyonların dan elde edilmiş farklı köken ve özelliklere sahip agregaların fiziksel özellikleri laboratuarda belirlenmiştir. Elde edilen agregalardan üretilen betonların basınç dayanımları laboratuar deneyleri ile tespit edilmiştir. Bu deney sonuçları geliştirilecek modelde girdileri oluşturacağı için düzenlenmiş ve kullanılabilir bir forma sokulmuştur. Betonun dayanım özelliklerinin belirlenmesi için daha önce yapılmış olan deneysel çalışmalardan yararlanılarak oluşturulan değişik yöntemler de kullanılmaktadır. Bu çalışmada İstanbul Anadolu yakasında değişik 10 ayrı ocaktan elde edilmiş agregaların fiziksel özellikleri, yapılan deneylerle belirlenmiştir. Fiziksel özellikleri belirlenen agregalarla beton örnekleri hazırlanmış ve bu beton örneklerin 7 ve 28 günlük basınç dayanımları ölçülmüştür. Betonu oluşturan agrega dışındaki bütün beton bileşenleri sabit tutularak değişen agregalarda betonun basınç dayanımı izlenmiştir. Deneysel olarak belirlenen değerlerin kestirimi için, Yapay Sinir Ağları Yöntemi kullanılarak modeller geliştirilmiş ve elde edilen sonuçlar karşılaştırılmıştır. YSA tekniği kullanılarak yeni bir model geliştirilmiş ve bu modelle betonun basınç dayanımını agreganın fiziksel özelliklerinden yola çıkılarak kestirilmesi amaçlanmıştır. Oluşturan modelde agreganın fiziksel ve mekanik özellikleri 21 farklı parametre ile temsil edilmiştir. Agrega özelliklerini belirleyen bu parametreler yapay sinir ağı modelinde giriş parametreleri olarak kullanılmıştır. Dolayısı ile 7 günlük ve 28 günlük beton dayanımlarını belirleyen YSA giriş katmanı 21 YSA hücresinden oluşturulmuştur. YSA çıkışında çıkış parametresi olarak yalnızca beton dayanımı bulunduğundan çıkış katmanında yalnızca bir hücre oluşturulmuştur. Geliştirilen modelle deney yapmadan agrega özelliklerinin girilmesi ile elde edilecek betonun basınç dayanımı tahmin edilebildiği gösterilmiştir. YSA modeli ile tahmin sonuçları ayrıca lineer regresyon yöntemi ile elde edilen sonuçlarla karşılaştırılmıştır. YSA sonuçları ile, deneysel veriler karşılaştırıldığında; YSA sonuçlarının % 2.8 gibi küçük bir hata oranıyla deneysel sonuçlarla uygunluk gösterdiği görülmüştür. YSA sonuçları ile lineer regresyon sonuçlarından daha başarılı olduğu ve %97’lere varan yakınlıkta tahminin gerçekleştiği gözlenmiştir

The effect of physical and mechanical properties of aggregate on concrete strength can only be determined with experiments. Experimental studies may take a long time to complete, create economic burden, require materials and technical personnel. Artificial Neural Network has shown us that these losses and requirements can be reduced. In this study, aggregate with different roots and properties were collected indifferent locations of Marmara region and their physical properties were specified. Acquired aggregate was used to create concrete and then their compressive strength was determined in laboratory experiments. The results of this experiment were organised and changed to a format that can be used as model input. Aggregate is one of the main raw materials used in the mixed concrete production. Therefore, aggregate quality carries a great importance for the mixed concrete producers. Grain size, grain shape, organic and alkali matter contents and mineralogical compositions are important material properties on the industrial evaluation of the aggregate deposits. Determination of the effect of the physical and mechanical properties of aggregates that have considerable effect on concrete strength is only possible by conducting a series of experimental studies.These studies take long time and mostly are not economic.Therefore, different methods formed by utilizing the experimental studies done before are used to determine the strength characteristics.In this study, the impact of physical properties of aggregates and using 7-day and 28-day cured concrete has been researched. Then a model has been developed by using artificial neural network which the results obtained from the tests. The values determined experimentally have been estimated by developing models in Artificial Neural Network method.It has been observed at the comparisons that the training and test results in the models can be estimated. Artificial Neural Networks are a family of statistical learning models inspired by biological neural networks (the central nervous systems of animals, in particular the brain) and are used to estimate or approximate functions that can depend on a large number of inputs and are generally unknown. Artificial neural networks are generally presented as systems of interconnected "neurons" which send messages to each other. The connections have numeric weights that can be tuned based on experience, making neural nets adaptive to inputs and capable of learning. Artificial Neural Networks are processing devices (algorithms or actual hardware) that are loosely modeled after the neuronal structure of the mamalian cerebral cortex but on much smaller scales. A large Artificial Neural Network might have hundreds or thousands of processor units, whereas a mamalian brain has billions of neurons with a corresponding increase in magnitude of their overall interaction and emergent behavior. Although Artificial Neural Network researchers are generally not concerned with whether their networks accurately resemble biological systems, some have. Examinations of humans' central nervous systems inspired the concept of artificial neural networks. In an artificial neural network, simple artificial nodes, known as "neurons", "neurodes", "processing elements" or "units", are connected together to form a network which mimics a biological neural network. There is no single formal definition of what an artificial neural network is. However, a class of statistical models may commonly be called "Neural" if it possesses the following characteristics: -Contains sets of adaptive weights, -Capability of approximating non-linear functions of their inputs. The adaptive weights can be thought of as connection strengths between neurons, which are activated during training and prediction. Neural networks are similar to biological neural networks in the performing of functions collectively and in parallel by the units, rather than there being a clear delineation of subtasks to which individual units are assigned. The term "neural network" usually refers to models employed in statistics, cognitive psychology and artificial intelligence. Neural network models which emulate the central nervous system are part of theoretical neuroscience and computational neuroscience. The word network in the term 'artificial neural network' refers to the inter–connections between the neurons in the different layers of each system. An example system has three layers. The first layer has input neurons which send data via synapses to the second layer of neurons, and then via more synapses to the third layer of output neurons. More complex systems will have more layers of neurons, some having increased layers of input neurons and output neurons. The synapses store parameters called "weights" that manipulate the data in the calculations. An Artificial Neural Networks is typically defined by three types of parameters: -The interconnection pattern between the different layers of neurons -The learning process for updating the weights of the interconnections -The activation function that converts a neuron's weighted input to its output activation. Training a neural network model essentially means selecting one model from the set of allowed models that minimizes the cost criterion. There are numerous algorithms available for training neural network models; most of them can be viewed as a straightforward application of optimization theory and statistical estimation.Most of the algorithms used in training artificial neural networks employ some form of gradient descent, using backpropagation to compute the actual gradients. This is done by simply taking the derivative of the cost function with respect to the network parameters and then changing those parameters in a gradient-related direction. The backpropagation training algorithms are usually classified into three categories: steepest descent, quasi-Newton and conjugate gradient. In this study, aggregate samples had been taken from 5 different Gebze rock quarries and 5 different Ömerli quaries which is in Anatolian site of İstanbul. In this study two different prediction models have been developed and compared to determine crushed aggregate concrete compressive strength using Regression Technique and Artificial Neural Network method. Experimental data which obtained from aggregate concretes used for developing both Artificial Neural Network and Regression Technique models. While developing the models to predict of compressive strength of concretes aggragates physical properties used. Predicted concrete compressive strengths using developed models compared with the experimental compressive strengths so the reliability of models tested. A new model was developed using Artificial Neural Network technique and this technique aims to calculate concrete compressive strength from aggregates physical properties. In the created model, 21 different parameters are used to identify aggregates physical and mechanical properties. These parameters that specify aggregate properties are used as input in artificial neural network. As a result, Artificial Neural Network input layer consists of 21 Artificial Neural Network cells when determining 7day and 28 day concrete strength. In Artificial Neural Network output, only given output parameter is concrete strength and this results in a single cell output. The developed model indicates that without making an experiment, providing input on properties of aggregate allows us to estimate concrete compressive strength. Artificial Neural Network model estimations and linear regression results are also compared. According to this comparison between Artificial Neural Network results and experiment data; it was seen that Artificial Neural Network result has a really small error margin of 2.8% percentage in difference. Artificial Neural Network results are observed to be more successful then linear regression results and estimations are close to 97%.