Elektrik Elektronik Mühendisliği Bölümü Koleksiyonu
Bu koleksiyon için kalıcı URI
Güncel Gönderiler
Öğe Determining of gas type in counter flow vortex tube using pairwise fisher score attribute reduction method(Elsevier Sci Ltd, 2011) Polat, Kemal; Kırmacı, VolkanThis paper focused on the determining of gas types in counter flow type vortex tubes. In the present study, four different gas types including air, oxygen, nitrogen, and argon in the vortex tube with different inlet pressures and nozzle numbers have been used. The main aims of this paper are to investigate the correlations between gas types and input parameters comprising nozzle numbers, inlet pressures, inlet mass flow rate, temperature of cold outlet, temperature of hot outlet, and cold mass fraction and to select the most important attributes using correlation based attribute reduction and pairwise fisher score attribute reduction (PFSAR). After attribute reduction methods applied to dataset, k-nearest neighbor and C4.5 decision tree classifiers have been used to determine the gas type in the RHVT. The results have demonstrated that the PFSAR is a robust and efficient method in the reduction of attributes belonging to vortex tube. (C) 2011 Elsevier Ltd and IIR. All rights reserved.Öğe Comparative analysis of transformer fault classification based on DGA data using machine learning algorithms(Ieee, 2024) Çoban, Melih; Fidan, Murat; Aytar, OktayDissolved gas analysis (DGA) is considered a leading technique for fault classification in power transformers. However, accurate analysis results can only be achieved if the measured gases are interpreted, appropriately. In DGA interpretation, traditional techniques, artificial intelligence techniques such as machine learning algorithms, and hybrid techniques are generally used. In this study, four well-known machine learning algorithms have been compared in terms of DGA fault classification: Support Vector Machine (SVM), K-Nearest Neighbor (KNN), Naive Bayes (NB) and Decision Tree (DT). The lowest accuracy rate was obtained as 63.63% using the NB algorithm and raw data. In addition to raw data, data converted to logarithmic form has been also used to develop classification models. The highest accuracy rate was determined as 94.54% using the DT algorithm and logarithmic data. The obtained results have been demonstrated the efficiency and stability of the DT algorithm for transformer fault classification, especially when the data was appropriately preprocessed.Öğe Multimodal joint prediction of traffic spatial-temporal data with graph sparse attention mechanism and bidirectional temporal convolutional network(Elsevier Sci Ltd, 2024) Zhang, Dongran; Yan, Jiangnan; Polat, Kemal; Alhudhaif, Adi; Li, JunTraffic flow prediction plays a crucial role in the management and operation of urban transportation systems. While extensive research has been conducted on predictions for individual transportation modes, there is relatively limited research on joint prediction across different transportation modes. Furthermore, existing multimodal traffic joint modeling methods often lack flexibility in spatial-temporal feature extraction. To address these issues, we propose a method called Graph Sparse Attention Mechanism with Bidirectional Temporal Convolutional Network (GSABT) for multimodal traffic spatial-temporal joint prediction. First, we use a multimodal graph multiplied by self-attention weights to capture spatial local features, and then employ the Top-U sparse attention mechanism to obtain spatial global features. Second, we utilize a bidirectional temporal convolutional network to enhance the temporal feature correlation between the output and input data, and extract inter-modal and intra-modal temporal features through the share-unique module. Finally, we have designed a multimodal joint prediction framework that can be flexibly extended to both spatial and temporal dimensions. Extensive experiments conducted on three real datasets indicate that the proposed model consistently achieves state-of-the-art predictive performance.Öğe Spatio-temporal characterisation and compensation method based on CNN and LSTM for residential travel data(Peerj Inc, 2024) Alhudhaif, Adi; Polat, KemalCurrently, most traffic simulations require residents' travel plans as input data; however, in real scenarios, it is difficult to obtain real residents' travel behavior data for various reasons, such as a large amount of data and the protection of residents' privacy. This study proposes a method combining a convolutional neural network (CNN) and a long short-term memory network (LSTM) for analyzing and compensating spatiotemporal features in residents' travel data. By exploiting the spatial feature extraction capability of CNNs and the advantages of LSTMs in processing time-series data, the aim is to achieve a traffic simulation close to a real scenario using limited data by modeling travel time and space. The experimental results show that the method proposed in this article is closer to the real data in terms of the average traveling distance compared with the use of the modulation method and the statistical estimation method. The new strategy we propose can significantly reduce the deviation of the model from the original data, thereby significantly reducing the basic error rate by about 50%.Öğe Convolution smoothing and non-convex regularization for support vector machine in high dimensions(Elsevier, 2024) Wang, Kangning; Yang, Junning; Polat, Kemal; Alhudhaif, Adi; Sun, XiaofeiThe support vector machine (SVM) is a well-known statistical learning tool for binary classification. One serious drawback of SVM is that it can be adversely affected by redundant variables, and research has shown that variable selection is crucial and necessary for achieving good classification accuracy. Hence some SVM variable selection studies have been devoted, and they have an unified empirical hinge loss plus sparse penaltyformulation. However, a noteworthy issue is the computational complexity of existing methods is high especially for large-scale problems, due to the non-smoothness of the hinge loss. To solve this issue, we first propose a convolution smoothing approach, which turns the non-smooth hinge loss into a smooth surrogate one, and they are asymptotically equivalent. Moreover, we construct computationally more efficient SVM variable selection procedure by implementing non-convex penalized convolution smooth hinge loss. In theory, we prove that the resulting variable selection possesses the oracle property when the number of predictors is diverging. Numerical experiments also confirm the good performance of the new method.Öğe New approaches to epileptic seizure prediction based on EEG signals using hybrid CNNs(Inderscience Enterprises Ltd, 2024) Nour, Majid; Arabacı, Bahadır; Öcal, Hakan; Polat, KemalThis study employs the University of Bonn Dataset to address the importance of frequency information in EEG data and introduces a methodology utilising the short-time Fourier transform. The proposed method transforms conventional 1D EEG signals into informative 2D spectrograms, offering an approach for advancing the detection of neurological diseases. Integrating advanced CNN architectures with the conversion of EEG signals into 2D spectrograms forms the foundation of our proposed methodology. The 1D CNN model utilised in this study demonstrates exceptional performance metrics, achieving a specificity of 0.996, an overall test accuracy of 0.991, a sensitivity of 0.987, and an F1 score of 0.989. Shifting to the 2D approach discloses a slight reduction in accuracy to 0.987, sensitivity of 0.976, specificity of 0.988, and an F1 score of 0.97. This analysis provides nuanced insights into the performance of 1D and 2D CNNs, clarifying respective strengths in the context of neurological disease detection.Öğe Corn processing by pulsed electric fields with respect to microbial inactivation and improvement of seed vigour(Elsevier Sci Ltd, 2024) Evrendilek, Gülsün Akdemir; Atmaca, Bahar; Uzuner, SibelPulsed electric field (PEF) treatment of corn grains to improve seed vigour and inactivation of endogenous microflora by energies ranging from 1.20 to 28.8 J were applied to determine effectiveness of applied energies on germination rate (GR), normal seedling rate (NSR), electrical conductivity (EC), ability to germinate under salt (100- and 200 mM salt) and cold (at 10 degrees C for 7 days and at 25 degrees C for 5 days) stresses. Moreover, the effect of PEF treatments was further investigated for the inactivation of total aerobic mesophilic bacteria (TAMB), total mold and yeast (TMY), and inactivation rate (%) of Aspergillus parasiticus. Increased energy provided 11.10 % increase in GR, 21.22 % increase in NSR, 95.50 % increase in germination at 10 degrees C for 7 days. Germination under stress conditions revealed 32.53 %, 68.35 %, and 76 % increase in germination at 25 degrees C for 5 days, under 100 mM- and 200 mM NaCI salt stresses. Inactivation on the mean initial TAMB and TMY were approximately 9.25 and 7.93 log, respectively, with 63.33 +/- 0.22 % reduction in A. parasiticus culture. PEF treated corn seedlings had stronger and taller body formation with stronger roots. The most optimal processing parameters were detected as 300 Hz, 28.80 J, and 19.78 sec. PEF treatment carries a high potential to improve corn vigour with inactivation of surface microflora.Öğe Pulsed electric field processing of fruit juices with inactivation of enzymes with new inactivation kinetic model and determination of changes in quality parameters(Elsevier Sci Ltd, 2024) Evrendilek, Gülsün Akdemir; Özkan, Birsen HititPulsed electric fields (PEF) processing of fresh grape, orange and tomato juices by 17.2-24.1 kV/cm electric field strengths, 130.0-1040.5 mu s treatment times and 1.43-29.13 J/kg energies with respect to determination of the changes in the physical and bioactive properties, color parameters in addition to inactivation of pectin methyl esterase (PME), polyphenol oxidase (PPO), and lipoxygenase (LOX) enzymes were studied. In general, decrease in pH, increase in conductivity (except for grape juice), total phenolic substance content, and total antioxidant capacity with no significant change (p > 0.05) in total soluble solids were observed by PEF treatment for all juices. Except for hue value for orange juice, no significant changes (p > 0.05) were observed in color values. Inactivation of the enzymes were increased with increased energy, and calculated electrical activation energies for PME, PPO, and LOX enzymes were 8.31, 7.06, and 4.31 kJ/mol, respectively. Industrial relevance: Enzymes in addition microorganisms causing quality degradation is of great concern in fruit juice industry as most of the enzymes are resistant to processing technologies. Increased PEF treatment time provided over 95% inactivation on PME, PPO, and LOX enzymes with no adverse effect on quality properties of grape, orange, and tomato juices revealing a viable alternative to juice processing industry.Öğe Privileged information learning with weak labels(Elsevier, 2023) Xiao, Yanshan; Ye, Zexin; Zhao, Liang; Kong, Xiangjun; Liu, Bo; Polat, Kemal; Alhudhaif, AdiPrivileged information learning is proposed to construct the classifier by incorporating privileged knowledge. At present, most of the privileged information learning methods assume that the instance is accurately labeled. However, in real-world applications, an instance may be weakly labeled. In this paper, we propose a novel privileged information learning method with weak labels (PLWB). The hypothesis of our work is that an instance may be annotated by a number of labelers and different labelers may give different labels to this instance due to distinct professional knowledge and subjective factors. It leads to ambiguous labels of instances, namely weak labels. To solve this problem, our methodology is to give each labeler a weight and incorporate these weights into a privileged information learning model. Our technique is to employ a heuristic framework to optimize the labeler weights and the privileged information learning model jointly. The existing privileged information learning methods do not consider the weak label problem, and assign an equal or random weight to each labeler. Our work is different from these methods. The novelty and theoretical contribution is that this is the first work to deal with the weak label problem in privileged information learning. The merit is that we assign an unknown weight to each labeler and solve the optimal values of these weights in the optimization process, such that the performance of the learning model can be improved with the optimal labeler weights. In the experiments, the tool that we use is MATLAB, in which we implement our algorithm. The experimental datasets include one handwritten categorization dataset, two image classification datasets (i.e., Animals-with-Attributes dataset and Caltech-101 dataset), and one disease diagnosis dataset (i.e., Alzheimer's Disease Neuroimaging Initiative dataset), in which the number of instances used is 2000, 6180, 8677 and 202, respectively. The obtained results are that: (1) by optimizing the labeler weights, the proposed PLWB method obtains explicitly higher classification accuracy than the existing privileged information learning methods; (2) PLWB has relatively higher training time since it needs to solve the labeler weights in the optimization process.& COPY; 2023 Elsevier B.V. All rights reserved.Öğe Design and implementation of drive and control system for ultrasonic motor over power line communication(Taylor & Francis Inc, 2024) Daldal, Nihat; Aytar, Oktay; Bekiroğlu, Erdal; Bal, GüngörIn this study, remote control application of an ultrasonic motor (USM) has been achieved over the power line communication (PLC) system. Fast, practical, affordable and effective operating mode is essential for the USM. This study aimed to develop an original, efficient, effective and economical method. Drive and control of USM control has been succeeded with the developed PLC control system. A two-phase high-frequency inverter, a power line transmitter, and a power line receiver circuits have been designed to drive and control of the ultrasonic motor. Required measurements are acquired from the power line to select the most suitable communication frequency and coupling circuit impedance for the PLC system. For the communication frequency and impedance value measurements the receiver and the transmitter circuits have been designed. The PLC-controlled system has been tested for different operating conditions of the ultrasonic motor. USM control has been accomplished over the existing power line without using extra cables and interfaces for communication. The obtained results show that the PLC-controlled system is practical, reliable, cost-effective, and feasible for the remote control of the USM. This research contributes a new and essential perspective for the PLC-based remote control studies in addition to the USM drive and control strategies.Öğe Dynamic trajectory partition optimization method based on historical trajectory data(Elsevier, 2024) Yu, Xiang; Zhai, Huawei; Tian, Ruijie; Guan, Yao; Polat, Kemal; Alhudhaif, AdiPartitioning dynamic trajectory data can improve the efficiency and accuracy of trajectory data processing, provide a foundation for trajectory data mining and analysis. However, with the continuous growth of trajectory data scales and the urgent demand for trajectory query efficiency and accuracy, partitioning methods have become crucial. The partitioning method of dynamic trajectory data faces significant challenges in terms of spatiotemporal trajectory locality, partition load balancing, and partition time. To address these challenges, we propose a method based on historical trajectory pre-partitioning, which can store data more effectively in distributed systems. We partition similar historical trajectory data to achieve preliminary partitioning of the data. In addition, we also construct a cost model to ensure that the workload of each partition is close to consistency. Extensive experiments have demonstrated the excellent partitioning efficiency and query efficiency achieved by our design compared to other partitioning methods.Öğe Measurement and evaluation of solar panel data via dc power line(IEEE, 2022) Daldal, Nihat; Uzun, Berat; Bekiroğlu, ErdalToday, it is important to monitor the data first of all in order to increase the efficiency of solar panels. In this study, parameters affecting the efficiency of photovoltaic panels, such as ambient temperature, panel temperature, humidity, light ratio, panel current, panel voltage were measured at certain time intervals. The data was then transferred to the computer via the PV panel (Power Line Communication, PLC) using the existing photovoltaic panel DC power line with the FSK modulator-demodulator, which was converted into a serial information package and designed, and the data was recorded. Here the panel data is collected entirely via its own energy cable without the use of any lines or wireless units. With Python-based software, graphs of panel parameters were created, the data obtained were analyzed and the factors affecting energy production were examined.Öğe Brain tumor detection with multi-scale fractal feature network and fractal residual learning(Elsevier, 2024) Jakhar, Shyo Prakash; Nandal, Amita; Dhaka, Arvind; Alhudhaif, Adi; Polat, KemalDeep learning has enabled the creation of several approaches for segmenting brain tumors using convolutional neural networks. These methods have come about as a direct result of the advancement of the field of machine learning. The proposed pixel-level segmentation is based on fractal residual deep learning; provide an insufficient degree of sensitivity when used for tumor segmentation. This is achieved due to fractal feature extraction and multi-scale approach used for segmentation. If multi-level segmentation is used, it is possible to effectively increase the sensitivity of the segmentation process which is the additional benefit from the proposed method. In this work, the production of tumor region is based on multi-scale pixel segmentation. This approach protects the integrity of tumor information while simultaneously improving the detection accuracy by cutting down on the total number of tumor regions. When compared to the information about the brain found in tumor locations, the proposed strategy has the potential to enhance the percentage of brain tumor information. This work proposes a novel network structure known as the Mutli-scale fractal feature network (MFFN) to increase the accuracy of the network's classification as well as its sensitivity when it comes to the segmentation of brain tumors. The proposed method with overall feature results in 94.66% accuracy, 94.42% sensitivity and 92.81% specificity using 5fold cross validation. In this paper the Cancer Imaging Archive (TCIA) dataset has been used in order to evaluate performance evaluation metrics and segmentation results to quantify the superiority of proposed brain tumor detection approach in comparison to existing methods.Öğe Detection of Alzheimer's disease from EEG signals using explainable artificial intelligence analysis(Institute of Electrical and Electronics Engineers Inc., 2024) Arabacı, Bahadır; Öcal, Hakan; Polat, KemalIn this study, the evaluation of classification models with frequency and chaotic features was aimed for the classification of healthy individuals and Alzheimer's patients using EEG signals. Morlet wavelet transform was employed for calculating EEG features to determine the characteristics in the frequency domain. Additionally, Lyapunov exponents were utilized for the analysis of chaotic features, and significant EEG channels were identified from the obtained results of the wavelet transform. Using permutation importance, the impact of each feature on the performance of the classification model was assessed. In this evaluation, the Random Forest model stood out in overall performance, showing the highest accuracy (0.7614), precision (0.7546), and F1 score (0.793) compared to other models. Furthermore, the Naive Bayes model achieved the highest sensitivity (0.8662) in detecting positive instances. © 2024 IEEE.Öğe Message from the general chair(Institute of Electrical and Electronics Engineers Inc., 2023) Polat, KemalThese proceedings contain papers presented at 2023 Asia-Europe Conference on Electronics, Data Processing and Informatics (ACEDPI 2023) was taken place in Prague, Czech Republic on April 17-19, 2023. ACEDPI 2023 aims to bring together leading academic scientists, researchers and research scholars to exchange and share their experiences and research results on all aspects of electronics, data processing and informatics.Öğe Classification of freezing of gait in Parkinson's disease using machine learning algorithms(Institute of Electrical and Electronics Engineers Inc., 2023) Önder, Mithat; Şentürk, Ümit; Polat, Kemal; Paulraj, D.Freezing of gait (FoG) is a prevalent and incapacitating symptom that affects individuals diagnosed with Parkinson's disease (PD) and other movement disorders. Detecting FoG is crucial for accurate diagnosis, fall prevention, and providing objective measurements, all of which are essential for optimizing treatment strategies and improving the quality of life for individuals with FoG. In this study, FoG has been detected using three different classification algorithms: Medium Gaussian Support Vector Machine (SVM), Medium K-Nearest Neighbor (KNN), and Boosted Trees. The process starts with data segmentation, where the dataset is divided into smaller segments. Then, feature extraction is performed on each segment to obtain various statistical measures such as mean, root mean square, maximum, standard deviation, kurtosis, skewness, and peak of root mean square. To ensure a robust and reliable analysis, the dataset is resampled using bootstrapping, a statistical technique that involves drawing random samples from the dataset with replacement. This leads to a more representative sample and reduces the impact of outliers or imbalanced data. The next step is to split the resampled dataset into three different approaches for the classification algorithm: In 5-FCV, the dataset is divided into five equal-sized subsets. Similarly, 10-FCV splits the dataset into ten subsets and follows the same process. Finally, the Medium Gaussian SVM, Medium KNN, and Boosted Trees classification algorithms are applied to the FoG dataset. The classification accuracy achieved is 86.9%, 87.6%, and 92.7% with 10-fold cross-validation, indicating that these algorithms are effective in accurately classifying FoG. © 2023 IEEE.Öğe The investigation of CMOS inverter based comparator circuits(Institute of Electrical and Electronics Engineers Inc., 2018) Keleş, Fatmanur; Aytar, OktayIn this study, the performance of the CMOS inverter circuit with active load, which can be used as a comparator structure in analogue digital converter circuits, is investigated with respect to other CMOS inverter circuits by using 0.25?m CMOS technology library in Cadence Virtuoso 6.13 design program. As a result of the analyzes made, it is seen that the proposed structure only consumes 1,041 uW and the delay time is 36.11 ps. The power dissipation, which the proposed design consumes, is significantly lower than the other designed CMOS inverter circuits. Therefore, when the proposed architecture is used in the comparator block of high-speed parallel analogue digital converters, power consumption is likely to decrease. © 2018 IEEE.Öğe Real time heart rate detection using non-contact photoplethysmography signals(IEEE Computer Society, 2014) Kavsaoğlu, Ahmet Reşit; Polat, Kemal; Bozkurt, Mehmet RecepHeart is contracted rhythmically so as to drive nutrients and oxygen necessary for life through our organs with blood arteries. The frequency for the rhythmic contraction of heart just as a pump is called heart rate (HR). Heart rate variation (HRV) is a measure of a fluctuation of time interval between heart beats. HRV is calculated considering electrodiagram (ECG) signals, arterial blood pressure signals or photoplethysmography (PPG) signals-derived time series of in-between heart beats. HRV is used as a significant indicator for the detection of healthiness and sickness state. Such pathological cases as high blood pressure, heart failure, and septic shock can be diagnosed using HRV. Therefore, accurate and rapid detection of HR is essential to correct diagnosis. In this study, real-time heart rate detection was derived from contactless PPG signals. PPG calling for contact with skin becomes useless in case of tissue scars or burns. In such cases, the use of contactless PPG is superior. Contactless PPG consists of a light source and a camera that senses reflection or transmittance of the light source. Camera images obtained were processed through an interface prepared in the MATLABTM GUI setting, and real-time heart rate detection was carried out. © 2014 IEEE.Öğe Performance analysis of machine Learning algorithms on power quality disturbances classification(Institute of Electrical and Electronics Engineers Inc., 2024) Gümüş, Birsen; Çoban, Melih; Tezcan, Süleyman SungurThe importance of the concept of power quality in electrical power systems is increasing. This situation causes the acceleration of work on detecting and eliminating power quality events. It is possible to automatically detect and classify these events using signal processing techniques and machine learning systems. In this study, the performances of Decision Tree (DT), Support Vector Machine (SVM), K Nearest Neighbor (KNN) and Naive Bayes (NB) classification algorithms were compared among themselves and with the literature. Signals containing 15 single and multiple power quality events were used for classification. Discrete Wavelet Transform (DWT) technique was preferred to extract the features of the signals. The data sets obtained by feature extraction were divided into test data and training data with a ratio of 20%, 30%, 40%. Thus, the classifiers were trained and tested with data sets of different sizes and their performances were evaluated. The highest accuracy rate was obtained as 99.75% with DT when 20% of the data set of noiseless signals was used as test data. The lowest accuracy rate was obtained as 30.6% with KNN when 30% of the data set of signals with 10dB noise was used as test data. In addition, the performances of the classifiers were examined by performing a 5-fold cross validation test and the average accuracy rates obtained from this were compared with other studies in the literature. © 2024 IEEE.Öğe Improving fault ride through capability of DFIG with fuzzy logic controlled crowbar protection(Institute of Electrical and Electronics Engineers Inc., 2022) Bekiroğlu, Erdal; Yazar, Muhammed DuranDoubly fed induction generators (DFIG) are sensitive to intense disturbances in the grid. Various methods such as crowbar protection circuits are used to increase the fault ride-through (FRT) capability of DFIGs. Crowbar is one of the most important protection methods used to protect the rotor side converter, rotor windings, and capacitors in the wind energy conversion system (WECS) from the damage of overcurrent/overvoltage. Setting the timing of the crowbar switch excellently is significant for the stability of the grid and the efficiency of the WECS. In this study, a Fuzzy logic controller (FLC) block has been added to control the crowbar switch to improve the timing of the crowbar and improve the FRT capability of the WECS. The performance of DFIG-based WECS with FLC under symmetric voltage dip was investigated by using Matlab/Simulink environment at a constant wind speed of 12 m/s. DC bus voltage, crowbar current, the terminal voltage of DFIG, and electromagnetic torque results were obtained. The results of the crowbar with the proposed controller and the results of the crowbar without FLC were compared. The results revealed that the proposed system is reliable, proper, and prevailing for the grid-connected DFIG under grid fault conditions. © 2022 IEEE.
