LEE- Savunma Teknolojileri Lisansüstü Programı
Bu topluluk için Kalıcı Uri
Gözat
Yazar "Karalar, Anıl Burak" ile LEE- Savunma Teknolojileri Lisansüstü Programı'a göz atma
Sayfa başına sonuç
Sıralama Seçenekleri
-
ÖgeAn artificial neural network approach to predict the results of strain gauge measurements in the tensile testing of unidirectional laminated composites(Graduate School, 2023) Karalar, Anıl Burak ; Balkan, Demet ; 831180 ; Savunma Teknolojileri Ana Bilim Dalı / Savunma Teknolojileri Bilim DalıIn the area of material science, obtaining material properties by using test methods is an exhausting and an expensive process. Especially in tensile testing, The specimen preparation is time-consuming, and buying specimens is also costly. In order to obtain reliable outcomes, a significant number of test samples incur damage. Therefore, the cost of testing systems is hardly affordable for many researchers, especially in the composite material development process for defense technologies area. New methods in the material testing area are necessary to avoid the economic burden of testing. The main idea of this thesis is to introduce an Artificial Neural Network (ANN) model for predicting a correlation between the device used for testing, and the strain gauges. Furthermore, the strain data obtained from the strain gauges is employed to derive the stress-strain curves essential for determining the material properties. In this thesis, the ANN model predicts the stress-strain curve. Thus, the different algorithms are modeled, and compared to select best algorithm for predicting a stress-strain curve obtained from different tests. Tensile testing is a crucial method for evaluating the mechanical properties of laminated composites. In this study, Artificial Neural Networks (ANN) were employed to analyze and summarize the tensile test results of laminated composites. The ANN models were trained using a dataset consisting of input variables such as displacement (mm), axial force (N), thickness (mm), length (mm), stress (MPa), and strain calculated from the displacement measured by the extensometer, when the output parameter is strain gauge readings. The objective was to develop a predictive model that could accurately estimate these mechanical properties based on the given input variables. Through an iterative training process, the ANN models were able to learn the complex relationships between the input variables and the tensile test results. Once trained, the models could make predictions for unseen laminated composite samples, providing valuable insights into their mechanical behavior without requiring for extensive physical testing. The accuracy and reliability of the ANN models were assessed through various statistical measures such as relative error, mean absolute error, root mean square error, and coefficient of determination and correlation. The results indicated that the developed ANN models were capable of accurately predicting the tensile properties of laminated composites based on the provided input variables. The use of ANN in this study offers several advantages. It provides a faster and more cost-effective alternative to traditional experimental testing, as the models can quickly analyze large amounts of data and provide predictions in real-time. Additionally, ANN models can capture complex nonlinear relationships between the input variables and the tensile properties, which may be challenging to identify using traditional analytical methods.