selim soylu | Aksaray University (original) (raw)

Papers by selim soylu

World Congress on Electrical Engineering and Computer Systems and Science, 2016

Type 1 Diabetes Mellitus (T1DM) is a worldwide disease. Although a complete cure has not been fou... more Type 1 Diabetes Mellitus (T1DM) is a worldwide disease. Although a complete cure has not been found yet, an artificial pancreas (AP), also known as a closed-loop insulin therapy, is becoming more important for the treatment of this disease. Controller part of the AP can compute insulin infusion rate that will keep blood glucose concentration (BGC) in normoglycemic ranges for patients with T1DM. In this paper, three different control algorithms are proposed as a controller part of the AP. These control algorithms include genetic algorithm based proportional-integral-derivative (GA-PID) control, artificial bee colony algorithm based PID (ABC-PID) control, and particle swarm optimization algorithm based PID (PSO-PID) control. In silico control studies are implemented through a virtual diabetic patient based on the Stolwijk-Hardy's glucose-insulin regulation model. Simulations are performed to assess control function in terms of tracking BGC profile of a healthy person against to a daily food intake of three meals. In order to demonstrate robustness, sensor noise test is implemented. Simulation results are promising in terms of regulating the daily BGC.

Biocybernetics and Biomedical Engineering, 2018

Background and objectives: Despite therapeutic advances, a complete cure has not been found yet f... more Background and objectives: Despite therapeutic advances, a complete cure has not been found yet for patients with type 1 diabetes (T1D). Artificial pancreas (AP) is a promising approach to cope with this disease. The controller part of the AP can compute the insulin infusion rate that keeps blood glucose concentration (BGC) in normoglycemic ranges. Most controllers rely on model-based controllers and use manual meal announcements or meal detection algorithms. For a fully automated AP, a controller only using the patient's BGC data is needed. Methods: An optimized Mamdani-type hybrid Fuzzy P + D controller was proposed. Using the University of Virginia/Padova Simulator, a 36 h scenario was tested in nine virtual adult patients. To take into account the effect of continuous glucose monitor noise, the scenario was repeated 25 times for each adult. The main outcomes were the percentage of time BGC levels are in the euglycemic range and blood glucose risk index (BGRI), respectively. Results: The obtained BGC values were found to be in the euglycemic range for 82.6% of the time. Moreover, the BGC values were below 50 mg/dl, below 70 mg/dl and above 250 mg/dl for 0%, 0.35% and 0.74% of the time, respectively. The BGRI, low blood glucose index (LBGI), and high blood glucose index (HBGI) were also found as 3.75, 0.34 and 3.41, respectively. The proposed controller both increases the time the BGC levels are in the euglycemic range and causes less hypoglycemia and hyperglycemia relative to the published techniques studied in a similar scenario and population.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES, 2018

In this paper, a Mamdani-type fuzzy controller is proposed as the controller part of an artificia... more In this paper, a Mamdani-type fuzzy controller is proposed as the controller part of an artificial pancreas. The controller is optimized with the artificial bee colony optimization algorithm. The glucose-insulin regulatory system, based on a nonlinear differential model in the presence of delay, is used both for virtual patient and healthy person data. The main target of the controller is to mimic a blood glucose concentration profile of the healthy person with exogenous insulin infusion. Simulations are performed to assess the control function in terms of tracking the blood glucose concentration profile of the healthy person and minimizing errors. To show robustness, a group of three tests are implemented. These tests include unusual glucose intake, sensor noise, and uncertainty in the clearance rate parameter. The simulation results demonstrate that the adopted method is more effective than similar studies in the literature.

Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science

This study represents a comprehensive investigation of the performance of Artificial Hummingbird ... more This study represents a comprehensive investigation of the performance of Artificial Hummingbird Algorithm for parameter estimation for Polymer Electrolyte Membrane Fuel Cell. With this purpose, four commercial fuel cell systems which were widely preferred in the literature such as NedStack PS6 (Case-I), 250 W fuel cell stack (Case-II), Horizon 500 W (Case-III), and BCS 500 W (Case-IV) were chosen. In order to compare the performance of this algorithm, seven well-known optimization techniques including Artificial Bee Colony, Salp Swarm Optimization, Particle Swarm Optimization, Gray Wolf Optimization, Genetic Algorithm, Harris Hawks Optimization, and Whale Optimization Algorithm were used. The sum of the squared errors, computational speed, and statistical measurements were calculated for the performance comparison. In this context, the best SSE values were found as 2.06556, 5.25017, 0.02477, 0.01170 for Case-I, Case-II, Case-III, and Case-IV, respectively. The best standard deviati...

2013 8th International Conference on Electrical and Electronics Engineering (ELECO), 2013

World Congress on Electrical Engineering and Computer Systems and Science, 2016

Type 1 Diabetes Mellitus (T1DM) is a worldwide disease. Although a complete cure has not been fou... more Type 1 Diabetes Mellitus (T1DM) is a worldwide disease. Although a complete cure has not been found yet, an artificial pancreas (AP), also known as a closed-loop insulin therapy, is becoming more important for the treatment of this disease. Controller part of the AP can compute insulin infusion rate that will keep blood glucose concentration (BGC) in normoglycemic ranges for patients with T1DM. In this paper, three different control algorithms are proposed as a controller part of the AP. These control algorithms include genetic algorithm based proportional-integral-derivative (GA-PID) control, artificial bee colony algorithm based PID (ABC-PID) control, and particle swarm optimization algorithm based PID (PSO-PID) control. In silico control studies are implemented through a virtual diabetic patient based on the Stolwijk-Hardy's glucose-insulin regulation model. Simulations are performed to assess control function in terms of tracking BGC profile of a healthy person against to a daily food intake of three meals. In order to demonstrate robustness, sensor noise test is implemented. Simulation results are promising in terms of regulating the daily BGC.

Biocybernetics and Biomedical Engineering, 2018

Background and objectives: Despite therapeutic advances, a complete cure has not been found yet f... more Background and objectives: Despite therapeutic advances, a complete cure has not been found yet for patients with type 1 diabetes (T1D). Artificial pancreas (AP) is a promising approach to cope with this disease. The controller part of the AP can compute the insulin infusion rate that keeps blood glucose concentration (BGC) in normoglycemic ranges. Most controllers rely on model-based controllers and use manual meal announcements or meal detection algorithms. For a fully automated AP, a controller only using the patient's BGC data is needed. Methods: An optimized Mamdani-type hybrid Fuzzy P + D controller was proposed. Using the University of Virginia/Padova Simulator, a 36 h scenario was tested in nine virtual adult patients. To take into account the effect of continuous glucose monitor noise, the scenario was repeated 25 times for each adult. The main outcomes were the percentage of time BGC levels are in the euglycemic range and blood glucose risk index (BGRI), respectively. Results: The obtained BGC values were found to be in the euglycemic range for 82.6% of the time. Moreover, the BGC values were below 50 mg/dl, below 70 mg/dl and above 250 mg/dl for 0%, 0.35% and 0.74% of the time, respectively. The BGRI, low blood glucose index (LBGI), and high blood glucose index (HBGI) were also found as 3.75, 0.34 and 3.41, respectively. The proposed controller both increases the time the BGC levels are in the euglycemic range and causes less hypoglycemia and hyperglycemia relative to the published techniques studied in a similar scenario and population.

TURKISH JOURNAL OF ELECTRICAL ENGINEERING & COMPUTER SCIENCES, 2018

In this paper, a Mamdani-type fuzzy controller is proposed as the controller part of an artificia... more In this paper, a Mamdani-type fuzzy controller is proposed as the controller part of an artificial pancreas. The controller is optimized with the artificial bee colony optimization algorithm. The glucose-insulin regulatory system, based on a nonlinear differential model in the presence of delay, is used both for virtual patient and healthy person data. The main target of the controller is to mimic a blood glucose concentration profile of the healthy person with exogenous insulin infusion. Simulations are performed to assess the control function in terms of tracking the blood glucose concentration profile of the healthy person and minimizing errors. To show robustness, a group of three tests are implemented. These tests include unusual glucose intake, sensor noise, and uncertainty in the clearance rate parameter. The simulation results demonstrate that the adopted method is more effective than similar studies in the literature.

Proceedings of the Institution of Mechanical Engineers, Part C: Journal of Mechanical Engineering Science

This study represents a comprehensive investigation of the performance of Artificial Hummingbird ... more This study represents a comprehensive investigation of the performance of Artificial Hummingbird Algorithm for parameter estimation for Polymer Electrolyte Membrane Fuel Cell. With this purpose, four commercial fuel cell systems which were widely preferred in the literature such as NedStack PS6 (Case-I), 250 W fuel cell stack (Case-II), Horizon 500 W (Case-III), and BCS 500 W (Case-IV) were chosen. In order to compare the performance of this algorithm, seven well-known optimization techniques including Artificial Bee Colony, Salp Swarm Optimization, Particle Swarm Optimization, Gray Wolf Optimization, Genetic Algorithm, Harris Hawks Optimization, and Whale Optimization Algorithm were used. The sum of the squared errors, computational speed, and statistical measurements were calculated for the performance comparison. In this context, the best SSE values were found as 2.06556, 5.25017, 0.02477, 0.01170 for Case-I, Case-II, Case-III, and Case-IV, respectively. The best standard deviati...

2013 8th International Conference on Electrical and Electronics Engineering (ELECO), 2013