Yazar "Ferati, Kajs" seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim Yok
    Öğe
    Artificial intelligence-based stress prediction in glass fiber reinforced composites
    (Korean Soc Mechanical Engineers, 2025) Ferati, Kajs; Adar, Nurettin Gokhan
    This study addresses the challenge of predicting stress values in glass fibre reinforced epoxy composites under tensile loading using advanced AI methods. Traditional experimental and finite element analysis (FEA) approaches are time consuming and costly. To overcome this, a dataset was generated from FEA simulations covering different lamination sequences and material properties. Three AI models-narrow neural network (NNN), squared exponential Gaussian process regression (GPR) and support vector machine (SVM)-were developed and evaluated. GPR and SVM achieved superior prediction accuracies of 96.83 % and 95.04 %, respectively, outperforming NNN. Experimental validation confirmed these results and demonstrated the robustness of the proposed models. This study provides a cost-effective framework for stress prediction in composites that reduces the reliance on extensive testing and simulation, and advances AI-driven solutions for materials design and analysis.
  • Küçük Resim Yok
    Öğe
    Prediction of Displacement and Stress Values of Composite Materials Under Load with Machine Learning Models
    (Osman SAĞDIÇ, 2022) Ferati, Kajs; Adar, Nurettin Gökhan
    In this study, the determination of displacement and stress values under certain load of glass fiber and epoxy resin laminated reinforced composite materials by using machine learning models is targeted. In the scope of study, the modelling is done by changing the material properties of varied laminations of composite samples via Ansys software and a tensile force is implemented in order to receive the total deformation and Von Misses stresses under the implemented tensile force and creation of the dataset is completed. The robust linear regression and Gaussian process regression models from machine learning algorithms are used to predict and determine the total deformation and Von Misses stresses by training and testing the models with the dataset created. As result, the predicted values obtained from trained and tested regression models and the real values obtained by modelling in Ansys are compared. Additionally, in consideration of model parameters for both regression models, the evaluation of true responses and correct prediction/determination is done. According to the results, Gaussian process regression model is determined as a better model for related study.