Altun, Kerem
Loading...
Profile URL
Name Variants
Altun, K.
Altun, K
Altun, K
Job Title
Email Address
keremaltun@iyte.edu.tr
Main Affiliation
03.10. Department of Mechanical Engineering
Status
Current Staff
ORCID ID
Scopus Author ID
Turkish CoHE Profile ID
Google Scholar ID
WoS Researcher ID
Sustainable Development Goals
1NO POVERTY
0
Research Products
2ZERO HUNGER
0
Research Products
3GOOD HEALTH AND WELL-BEING
0
Research Products
4QUALITY EDUCATION
2
Research Products
5GENDER EQUALITY
0
Research Products
6CLEAN WATER AND SANITATION
0
Research Products
7AFFORDABLE AND CLEAN ENERGY
2
Research Products
8DECENT WORK AND ECONOMIC GROWTH
0
Research Products
9INDUSTRY, INNOVATION AND INFRASTRUCTURE
3
Research Products
10REDUCED INEQUALITIES
0
Research Products
11SUSTAINABLE CITIES AND COMMUNITIES
0
Research Products
12RESPONSIBLE CONSUMPTION AND PRODUCTION
0
Research Products
13CLIMATE ACTION
0
Research Products
14LIFE BELOW WATER
0
Research Products
15LIFE ON LAND
0
Research Products
16PEACE, JUSTICE AND STRONG INSTITUTIONS
0
Research Products
17PARTNERSHIPS FOR THE GOALS
0
Research Products
Documents
17
Citations
954
h-index
7
WoS data could not be loaded because of an error. Please refresh the page or try again later.
Scholarly Output
6
Articles
0
Views / Downloads
2324/2199
Supervised MSc Theses
5
Supervised PhD Theses
0
WoS Citation Count
3
Scopus Citation Count
3
Patents
0
Projects
0
WoS Citations per Publication
0.50
Scopus Citations per Publication
0.50
Open Access Source
5
Supervised Theses
5
Journals data is not available
Scopus Quartile Distribution
Quartile distribution chart data is not available
Competency Cloud
6 results
Scholarly Output Search Results
Now showing 1 - 6 of 6
Master Thesis Error Analysis and Characterization of Piezoresistive Array Touch Sensors(01. Izmir Institute of Technology, 2022) Sarp, Mehmet Ogün; Altun, KeremNumerous investigations and academic studies employ piezoresistive types of sensors. One of the main foundations of this thesis is to contribute to the literature by making a detailed error analysis and by proposing a new error reduction. In this thesis, piezoresistive touch sensors are designed and manufactured from scratch in accordance with the working principle of commercially available sensors. This thesis examines two different sensor configurations for single-touch applications. On these sensors, three different loads are tested statically with the same 61 test points located on the sensor, a single load at a time, and their Center of Pressure results are examined and compared to each other. As a result, it is observed that the 7x7 sensor array gives more successful results than the 5x5 sensor array at the same test points. Kadane's algorithm is introduced and implemented in experiments aimed to reduce the error values. As a result, success is achieved. Furthermore, with another proposed method, circle fitting, the centers of the theoretical circle formed by measurements are found, and it is examined whether the sensor measurements were considered as homogeneous. In other words, in each case, the levels of decentralization did not vary much. Finally, the multivariate linear regression method is examined through the system equations obtained from the randomly selected measurement points. It is seen that both the sensor outputs of the other points on the system can be predicted and the error metrics on the system can be reduced.Master Thesis Wearable Systems for Performance Assessment in Volleyball(Izmir Institute of Technology, 2022) Özdemir, Muhammed Emin; Altun, KeremNowadays, wearable sensors are used for many applications such as healthcare, animation, sports, to name but a few. In this study, they are used to recognize volleyball activities such as digs, blocks, serves and spikes. These activities are normally followed by statisticians on the field, their presences and frequencies are noted by them to be recorded at the match report. This study focuses on automating this procedure and identifying/recognizing them using wearable sensors. Five Xsens MTw Awinda sensors are used to collect data from 10 volleyball players (5 women and 5 men) who are between 19-21 ages and have 3-12 years of experience as an active player in volleyball. In this thesis, optimum number of sensors and their locations, effects of combinations of different features such as minimum, maximum values, means and variances of the raw data, impacts of combinations of different sub sensors such as accelerometer, gyroscope and magnetometer on the 4-class&10-class classification average accuracies are investigated. Two classification algorithms are applied with two different cross validation methods: For both cross validation methods, LDA (Linear Discriminant Analysis) produced better average accuracies than KNN (K Nearest Neighbor) where k value is taken as 5. The average accuracies for 4-class and 10-class classifications are respectively 99.56% and 89.56%. However, these results are respectively 92.39% and 66.08% for KNN (k=5).Master Thesis Learning of Tasks With Robot Programming by Demonstration(Izmir Institute of Technology, 2022) Argüz, Serdar Hakan; Altun, Kerem; Ertuğrul, ŞenizIncreasingly more unstructured environments of today’s industry challenge the robots to have the capability to dynamically adapt to variations in the part sizes and positions. Traditional programming methods fall short of answering such needs. Programming by demonstration is an approach that allows the robots to learn tasks from human demonstrations. Improvements in the generalization of individual tasks that compose the complex assembly operations are an indispensable need for a more extensive adoption of PbD in the industry. This thesis aims to improve the generalization of the peg-in-hole task against variations in the hole positions. It uses the change in the hole position a metric for the novelty of the task and tests the success rate at increasing distances. The relationship between the novelty of the task and its success is examined for two different learning strategies. In the first strategy, only the positional characteristics of the task are learned, whereas both positional and force characteristics are learned in the latter. It is found that the success rate of the task decreases in both cases as the distance increases. However, the hybrid position/force learning strategy outperforms the purely positional one at all distances. As a result, this strategy is experimentally shown to be a valid approach to improve the generalization of the peg-in-hole task for changing hole positions. Incorporation of this strategy with existing frameworks and orientation generalization methods is suggested as future work.Master Thesis Touch Gestures Classification by Deep Learning Methods(Izmir Institute of Technology, 2022) Ege, Irmak; Altun, KeremIn this study, we carried out social touch gesture classification on two publicly available datasets, Corpus of Social Touch (CoST) and Human-Animal Affective Robot Touch (HAART), and our demo dataset. In order to classify touch gesture datasets, four different models are proposed: 3-dimensional convolutional neural network (3D-CNN), 3-dimensional convolutional-long term short term memory neural network (3D-CNNLSTM), 3-dimensional convolutional-bidirectional long term short term memory neural network (3D-CNN-BiLSTM) + and 3-dimensional convolutional transformers network (3D-CNN-Transformer). The fundamental layer of the proposed deep neural network architectures is 3-dimensional convolution layer that enables to extract spatio-temporal features of touch gestures. In this regard, with the use of spatio-temporal features of touch gestures, generalization performance of proposed four models have been improved using data augmentation techniques by applying randomly shift and rotation, and ensemble learning. Additionally, We also found out that Stochastic Gradient Descent (SGD) optimization algorithm has better generalization performance than Adaptive Moment Estimation (ADAM), which is used more frequently in deep learning. The accuracy of classification results of three dataset is investigated in terms of proposed model. The results showed that the proposed methods, especially ensemble classifier and the ensemble classifier with data augmentation, are beneficial for obtaining more generalizable learning algorithms. The scripts of deep neural network architecture are available upon request.Conference Object Citation - WoS: 3Citation - Scopus: 3Experimental Evaluation of the Success of Peg-In Tasks Learned From Demonstration(IEEE, 2022) Argüz, Serdar Hakan; Ertuğrul, Şeniz; Altun, KeremIndustrial robots are traditionally programmed by hard-coding the desired motion into them. That approach, however, costs significant time and effort and shows little to no promise in transferring human skills to robots. Programming by demonstration (PbD) is an alternative approach that allows robots to learn tasks from demonstrations. Because of its several advantages over the traditional method, PbD is particularly suited for tasks encountered in assembly operations, the most typical of which is the peg-in-hole task. A successful PbD implementation for a peg-in-hole task requires that the peg should still be inserted into the hole even under situations that are not encountered during the demonstrations. Previous research in the field shows that the success rate of a peg-in-hole task under such cases varies greatly. In this study, we use a UR5 manipulator to experimentally investigate how the success rate of a peg-in-hole task changes with respect to the novelty of the task, quantified in terms of the distance of the hole to its original position. It is found that the success ratio decreases as the novelty of the task increases. To increase the performance, the use of strategies that alter the robot's motion dynamically in the run time is suggested for future work.Master Thesis Modellıng And Sımulatıon Of A Doubly Fed Inductıon Generator-based Wındturbıne Under Symmetrıcal Voltage Fault(2023) Çiçek, Elif Dilara; Akkurt, Gülden Gökçen; Altun, KeremWhen wind turbines, which convert wind energy, one of the renewable energy sources, into electrical energy, are connected to the electrical grid, it is of great importance to maintain grid stability. However, the variable nature of wind can pose certain challenges to system stability in wind turbine grid integration. Regulations in different countries require wind turbines to continue contributing to the grid in the event of a fault. In this thesis, the process of converting wind energy into electrical energy through a specific generator is described, and the behaviour of the turbine is simulated in the event of a symmetrical voltage fault. This thesis aims to develop a simulation model for a variable-speed doubly fed asynchronous generator-based (Dfig) wind turbine with a partially scaled frequency converter control using the field-oriented vector control method and to investigate its grid contribution ability under different conditions. The aim was to ensure torque control on the generator side and grid-side control. After the completion of the control system, the turbine's grid contribution ability was examined under short-term faults, and it has been shown that the system continues to contribute to the grid after the voltage drop. The modelling was performed using the Matlab / Simulink program, and the results were shared. In the last part of the thesis, the effect of a symmetrical fault on the rotor current is analysed by the Monte Carlo method.