Gazi Akgun

Dr. Gazi Akgun

Marmara University, Istanbul, İstanbul, Turkey (Türkiye)
orcid

Research Focus & Metrics

Academic Career Length

9Years (2018-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 April 2024 in Journal of Thermoplastic Composite Materials
Verified
Journal Article
Closed access
Gazi AkgunGazi Akgun
Gazi Akgun
Osman UlkirOsman Ulkir
Osman Ulkir
The final product of additive manufacturing (AM) or 3D printing critically depends on the surface quality. An experimental study on the 3D printed intake manifold flange using acrylonitrile butadiene styrene (ABS) material was executed by varying the four process parameters. A fused deposition modeling (FDM) based 3D printer was used to fabricate the flanges. The association between the parameters and the surface roughness of printed ABS flanges was investigated. A feed forward neural network (FFNN) model trained on particle swarm optimization (PSO) optimized with a genetic algorithm (GA) was used to estimate the surface roughness. A Box-Behnken design (BBD) with printing parameters at three levels was used, and 25 parts were fabricated. The suggested model demonstrated a coefficient of determination (R2) of 0.9865 on test values, mean of root-mean-square-error (RMSE) of 0.1231 after 500 times training for generalization. And also mean of overfitting factor is 0.7110. This means that the suggested system could generalize. Comparing the results from the suggested model and ANN, the suggested hybrid model outperformed ANN in predicting the surface roughness values with no overfitting. This suggests that GA optimized PSO based FFNN may be a more suitable method for estimating product quality in terms of surface roughness.

30 September 2019 in Transactions of the Institute of Measurement and Control
Verified
Journal Article
Closed access
PDF file
Gazi AkgunGazi Akgun
Gazi Akgun
ACAhmet Emre Cetin
Ahmet Emre Cetin
EKErkan Kaplanoglu
Erkan Kaplanoglu
An adaptive robotic system has been developed to be used for hand rehabilitation. Previously developed exoskeletons are either very complex in terms of mechanism, hardware and software, or simple but have limited functionality only for a specific rehabilitation task. Some of these studies use simple position controllers considering only to improve the trajectory tracking performance of the exoskeleton which is inadequate in terms of safety and health of the patient. Some of them focus only on either passive or active rehabilitation, but not both together. Some others use EMG signals to assist the patient, but this time active rehabilitation is impossible unless different designs and control strategies are not developed. The proposed mechanical structure is extremely simple. The middle and the proximal phalanxes are used as a link of consecutively connected two 4-bar mechanisms, respectively. The PIP and MCP joints are actuated by a single electro mechanical cylinder to produce complex flexion and extension movements. It is simpler than similar ones from aspect with the mechanical structure and the biodynamic fit of the hand, making it practicable in terms of production and personal usage. Simple design lets to implement adaptive compliance controller for all active and passive rehabilitation tasks, instead of developing complex and different strategies for different rehabilitation tasks. Furthermore, using the Luenberger observer for unmeasured velocity state variable, an on-line estimation method is used to estimate the dynamic parameters of the system. This makes possible to estimate the force exerted by the patient as well, without a force sensor.

27 May 2025 in Smart Wearable Technology
Verified
Journal Article
Open access
PDF file
EEErdem Erdemir
Erdem Erdemir
EKErkan Kaplanoglu
Erkan Kaplanoglu
CUCihan Uyanik
Cihan Uyanik
Gazi AkgunGazi Akgun
Gazi Akgun
Predicting the final grasp tendency at the start of movement in prosthetic hands is crucial for improved control. Biological data, such as 3D movement and muscle activity, have been using by researchers to predict the final grasp. Early prediction of the intended grasp allows the prosthetic device to initiate control actions before the motion is complete, resulting in faster and more intuitive responses. Most machine learning algorithms are trained to predict the gesture of the final grasp. The aim of this study is to accurately estimate the final grasp state using inertial measurement unit (IMU) data. This estimation, based on movement trajectories, will allow prosthetic devices to respond more quickly to user actions. Deep Learning model was trained using movement data collected from a prosthetic hand controlled certain gesture trajectories without any human involvement. Data such as acceleration, angular velocity, and orientation were gathered through IMU sensors to create 3D orientation matrices representing the movement process. A deep convolutional neural network was used for training, with data labeled by the final grasp states. The deep learning algorithm successfully predicted the final hand motion with 93% accuracy. This trained model enables the generation of smooth supervisory trajectories, facilitating faster and more accurate control of the prosthesis. The proposed model demonstrates significant potential in improving prosthetic hand control by predicting the final hand movement at an early stage of motion, contributing to more responsive and effective prosthetic devices. Received: 12 March 2025 | Revised: 24 April 2025 | Accepted: 12 May 2025 Conflicts of Interest The authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in GitHub respiratory at https://github.com/BioAstLab/MotionTrajectoryEstimation. Author Contribution Statement Erdem Erdemir: Conceptualization, Methodology, Software, Investigation, Resources, Writing – review & editing, Visualization, Supervision, Project administration. Erkan Kaplanoglu: Conceptualization, Methodology, Validation, Formal analysis, Investigation, Data curation, Writing – original draft, Visualization, Project administration. Cihan Uyanik: Software, Validation, Formal analysis, Resources, Data curation, Writing – review & editing. Gazi Akgun: Validation, Data curation, Writing – original draft.

1 October 2023 in Science and Technology of Welding and Joining
Verified
Journal Article
Closed access
Osman UlkirOsman Ulkir
Osman Ulkir
Gazi AkgunGazi Akgun
Gazi Akgun
The selection of parameters affects the surface roughness in the additive manufacturing process. This study aims to determine the optimal combination of input parameters for predicting and minimising the surface roughness of samples produced by Fused Deposition Modelling on a 3D printer using a cascade-forward neural network (CFNN) and genetic algorithm. Box–Behnken Design with four independent printing parameters at three levels is used, and 25 parts are fabricated with a 3D printer. Roughness tests are performed on the fabricated parts. Models generated by the hybrid algorithm achieve the best results for predicting and optimising surface roughness in 3D-printed parts. The surface roughness prediction accuracy of the trained CFNN with optimised parameters is more accurate compared to previous random test results.

15 May 2025 in Journal of Thermoplastic Composite Materials
Verified
Journal Article
Closed access
Osman UlkirOsman Ulkir
Osman Ulkir
Gazi AkgunGazi Akgun
Gazi Akgun
This study investigated the prediction of compressive strength in fused deposition modeling (FDM) printed acrylonitrile butadiene styrene (ABS) samples Taguchi-fuzzy logic (FL) approach. The research examined six critical printing parameters: printing direction (PD), infill density (ID), infill pattern (IP), layer height (LH), printing speed (PS), and nozzle temperature (NT). The Taguchi L27 orthogonal array was utilized to systematically design experiments, ensuring minimal trials while maximizing the amount of obtained data. Analysis of variance (ANOVA) revealed that ID was the most influential parameter with a 37.39% contribution to compressive strength, followed by PS (26.95%) and LH (17.51%). The optimal parameter combination achieved a maximum compressive strength of 55.76 MPa with 90% ID, 100 µm LH, and 40 mm/s PS. FL model was developed to predict compressive strength, demonstrating superior accuracy with a 3.1% average error rate compared to the Taguchi model’s 3.7%. The model’s reliability was validated through additional experiments, confirming its effectiveness for predicting compressive strength in FDM-printed ABS components. This research provides a robust methodology for accurately predicting mechanical properties, contributing to advancing additive manufacturing (AM) quality control.

9 October 2022 in Sensors
Verified
Journal Article
Open access
PDF file
Erkan KaplanogluErkan Kaplanoglu
Erkan Kaplanoglu
Gazi AkgunGazi Akgun
Gazi Akgun
This study proposed a control method, a data-driven predictive control (DDPC), for the hand exoskeleton used for active, passive, and resistive rehabilitation. DDPC is a model-free approach based on past system data. One of the strengths of DDPC is that constraints of states can be added to the controller while performing the controller design. These features of the control algorithm eliminate an essential problem for rehabilitation robots in terms of easy customization and safe repetitive rehabilitation tasks that can be planned within certain constraints. Experiments were carried out with a designed hand rehabilitation system under repetitive and various therapy tasks. Real-time experiment results demonstrate the feasibility and efficiency of the proposed control approach to rehabilitation systems.
1