Browsing by Author "AKSOY, ABDULLAH"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • No Thumbnail Available
    Publication
    Automatic soliton wave recognition using deep learning algorithms
    (Pergamon-Elsevier Science Ltd, 2023-07-17) Aksoy, Abdullah; Yiğit, Enes; AKSOY, ABDULLAH; YİĞİT, ENES; Bursa Uludağ Üniversitesi/Mühendislik Fakültesi/Elektrik-Elektronik Mühendisliği Bölümü.; AAH-3945-2021; JFJ-3503-2023
    In this study, deep learning (DL) based wave classification is performed to automatically recognize the soliton waves. Different experiments using non-linear transmission lines (NLTLs) are performed and the signal images obtained from the experiments are recorded. To demonstrate the applicability of the soliton wave in different scenarios, the waves are generated in different devices, under different noise conditions, and in various environments. Based on the images obtained from the experiments, four different classes consisting of sine, square, triangle, and soliton waves are created. 225 different images belonging to each classes are created and thus a total of 900 different image data are obtained. Five popular DL algorithms, namely DenseNet201, VGG16, VGG19, Xception, and ResNet152, are used to train and test the data. The DenseNet201 algorithm showed the best performance with 0.9904 training accuracy, 0.9630 validation accuracy, and 0.9778 test results. Thus, soliton waves are easily separated from other waveforms such as square, triangle, and sine. These results clearly demonstrate the feasibility of using DL algorithms to automatically recognize the soliton waves, which can have significant implications in various fields such as telecommunications, optics, nonlinear electronics, and nonlinear physics.
  • No Thumbnail Available
    Publication
    Soliton wave parameter estimation with the help of artificial neural network by using the experimental data carried out on the nonlinear transmission line
    (Pergamon-elsevier Science Ltd, 2023-02-15) AKSOY, ABDULLAH; YENİKAYA, SİBEL; Yenikaya, Sibel; Bursa Uludağ Üniversitesi/Mühendislik Fakültesi/Elektrik ve Elektronik Bölümü.; AAH-3945-2021
    In this study, an artificial neural network (ANN) model is generated, which is used to estimate the output pa-rameters of soliton waves produced as a result of nonlinear transmission lines (NLTLs). Three different output parameters are acquired as a consequence of the experiments carried out utilizing the five various input pa-rameters that are set in the ANN-based study. Input parameters for NLTL designs with 116 different experiments; inductor (L), input voltage (Vi) value, number of nodes (n), capacitance (C(V)) and load resistance (RLoad) values. Output parameters values, which are maximum voltage (Vmax), center frequency (fcenter), and voltage modulation depth (VMD). Input and output data; 70 % is set aside for training, 15 % for validation and the remaining 15 % for testing. Training, validation, and testing steps are repeated for the output parameters, in which case more than 99 % correlation is found as a result of each operation. An absolute percentage error value is found for each output parameter. Moreover, Mean absolute percentage error (MAPE) is calculated for these output datasets. The data set is tested for the experimental studies carried out in the literature, and it is observed that there is a compliance of over 99 % for this situation.