Funded Projects

Theses

 

Recent Theses


The Analysis of Characteristic Dynamics of Stock Management Structures
Stock management is a dynamic task which is often found in managerial, physical, and biological systems. The aim in stock management is to bring a stock at a desired level and maintain it at that level by taking corrective actions. Stock management task imposes difficulties to the decision maker, which results in unwanted oscillations. In this thesis, different stock management structures are modeled and analyzed. We carry out complete parametric analysis of stock management problems with continuous delays of first, second, and third orders aiming to obtain the range of values for different characteristic dynamics of stock. For parametric analysis, we use control theoretic approaches. We first provide different stock management structures modeled using stock-flow diagrams of system dynamics methodology. Secondly, we obtain the corresponding simplified differential equations of the system dynamics model and, based on the differential equations, we obtain block diagrams. Thirdly, we convert simplified differential equations of the model from time domain to s-domain using Laplace transformation technique and obtain the transfer function. Fourthly, the characteristic equation of the transfer function is determined. Finally, we determine the critical values of the decision parameters at which a qualitative change in dynamics is observed by analyzing the roots of the characteristic equation. The critical values that are reported in this thesis are valid for all durations of the delay between the corrective actions and their eventual results on the stock. We also obtained a few counterintuitive results such as increasing the level of aggressiveness in stock corrections can completely eliminate oscillations in one of the cases. Aiming to build a bridge between system dynamics and control theory, the corresponding block diagrams of many basic system dynamics models are provided in the thesis and also in one of its appendices.


Researcher:
Sema MEHMET
Contact:
hakan.yasarcan@boun.edu.tr

Behavior Analysis and Testing Software
Analysis of model behavior is mainly conducted in a pattern-based manner in system dynamics (SD) methodology. In pattern-based evaluation of model outputs, similarity of the overall behavior pattern (e.g. S-shaped-growth, oscillations) and of specific pattern characteristics (e.g. inflection points, periods, amplitudes) are more important than point-by-point similarity measures such as sum-of-squared errors. Although some output analysis tools/software that address this special pattern focus are available, they lack usability and are fragmented. In this study, new standalone analysis software, namely Behavior Analysis and Testing Software (BATS), is developed. It integrates a pattern classification algorithm and a set of statistical methods for analysis of steady-state behaviors. Apart from enabling comparison of behaviors with these algorithms/methods, BATS includes structured processes that enable user to conduct automated hypothesis testing, behavior space exploration, and sensitivity analysis. In its current state, BATS can seamlessly communicate with SD modeling software (Vensim) and other common data sources. This study provides illustrative examples of how BATS can assist the modeler and/or analyst in various phases of modeling; indirect structure testing, output evaluation, sensitivity analysis, policy analysis. Considering its pattern-orientation, user-friendly interface, and communication with modeling software BATS can be an important contribution to the analysis toolset of SD methodology.
Click here to go to the project webpage.


Researcher:
Can SÜCÜLLÜ
Contact:
can.sucullu@boun.edu.tr
gonenc.yucel@boun.edu.tr

Analyzing Electric Vehicle Diffusion Scenarios for Istanbul
In this study, a dynamic simulation model for electric vehicle (EV) diffusion is constructed. The objective of this work is to investigate two main questions; what are the plausible diffusion patterns of electric vehicles for Istanbul under different scenarios developed considering both local and global socio-economic, governmental, technological factors and their interaction with each other? Secondly, what is the extent of the diffusion rate that can be expected in Istanbul after three decades? The model is validated by standard structure and behavior tests. After, various scenario and policy analysis are performed. The results show that fleet market share of battery electric vehicle (BEV) and hybrid electric vehicle (HEV) would likely reach around 19.76% and 20.77% respectively by 2042 in Istanbul. In addition, CO2 reduction in the transportation sector would only reach around 17.32% in 2042. Moreover, both gasoline and electricity cost influence EV diffusion. However, their impact on EV diffusion is mainly related with a mobility cost gap between gasoline and electricity. Furthermore, technological improvement would lead BEV sales to increase. However, if battery technology cannot keep pace with or exceed CV technology, technological improvements would less likely create a significant raise in the BEV sales. Contrary to expectations, even if no technological improvements were progressed, BEVs would still likely succeed to penetrate around 10% of the market with its current technology within the 30 years. Moreover, a sufficient number of recharging points may lead to faster diffusion of BEV’s as well, causing higher fleet market share overall. Both marketing activities and word of mouth have a remarkable impact on rapid EV diffusion. Besides, increase in repurchasing rate may cause faster EV penetration. Subsidies would have a small impact on EV sales. Finally, applying the 3% private consumption tax (PCT) instead of the 37% PCT for HEV may increase HEV sales but does not show considerable change on HEV sales.


Researcher:
Özlem TURAN
Contact:
gonenc.yucel@boun.edu.tr

Decision Making Implications for a Selected Echelon in the Beer Game
The beer production-distribution game, in short “The Beer Game”, is essentially a board game and it simulates a four echelon supply chain consisting of a retailer, wholesaler, distributor, and factory. During the game, every participant in a group of four is responsible for one of these four echelons and manages the associated inventory by placing orders. The aim of the game is to minimize the accumulated total cost obtained by the participants of a group managing each echelon. In this thesis, a mathematical model that is an exact one-to-one replica of the board version is constructed. The main aim of this thesis is to develop an understanding about how one should control an echelon in The Beer Game in the presence of identically controlled echelons; we assume that only the participant managing the echelon of concern behaves different than the rest of the group. We are specifically interested in the case where the echelons other than the selected one sub-optimally manage their individual inventories or backlogs. There can be two objectives: (i) the minimum cost for the echelon of concern can be obtained, (ii) the minimum group total cost can be obtained by optimizing the decision parameters of the selected echelon. Accordingly, we optimize the parameters of the anchor-and-adjust heuristic, which is the control policy used in this study, for the selected echelon by keeping the decision parameters constant for the rest of the three positions. We obtain different instances of the anchor-and-adjust ordering policy by optimizing stock adjustment time and by optimizing desired inventory of the selected echelon. In general, the group total cost can be decreased by allowing an increase in the total cost of the selected echelon. Unexpectedly, we obtained the lowest group total costs for the wholesaler when we minimized the group total cost by sacrificing the objective of minimizing the cost of the echelon of concern.


Researcher:
Mert EDALI
Contact:
hakan.yasarcan@boun.edu.tr

Interactive Dynamic Modelling for the Passenger Flow Bottleneck and Security Check-Point Management in Atatürk Airport
In this project, an improvement in management of passenger flow in Atatürk Airport is aimed. With the contributions of TAV, the current structure of the passenger flows is examined. As a consequence, it is decided that security checkpoints are the main reasons of the bottlenecks in the passenger flows. In this regard, passenger flow and relation of security check points/personnel with this passenger flow was modeled dynamically. The model is set up based upon two main flows which are through the domestic and the international terminals. After analyzing real hourly flow data of both international and domestic terminals and figuring their statistical features out, results are used in the model as input data. Time unit of the model was determined as a minute and simulation time is decided as one day. After the simulation, instant number of active x-rays devices/personnel needed was obtained both graphically and numerically. After obtaining these numbers, necessary number of active x-ray/personnel was tried to distribute to each security points through a developed algorithm in a day. Validation and verification of the model is tested under various extreme conditions and different scenarios. Finally, with the purpose of making trial runs on the model, a game version of this model was built. At the end of this project, it is expected that by using this model and the simulation game, decision making structure of security personnel allocation is improved.


Researchers:
M. Serdar ÇAKACIASLAN
Tolga SAĞLIK
Pınar SEKE
Mustafa UĞUR
Contact:
serdar.cakaciaslan@boun.edu.tr

Dynamic Impacts of Performance Based Payment System on Public Hospitals in Turkey
The goal of pay for performance (P4P) system in healthcare is to increase the efficiency of healthcare resources by paying physicians and hospitals for performance. Ministry of Health in Turkey has implemented P4P since 2004. The purpose of this study is to investigate the dynamic impacts of P4P on the behaviors of hospitals and physicians. The model includes physicians’ interactions with patients, the revenue pressures on physicians, and the resulting impacts on health outputs and quality. In order to increase productivity, physicians are induced to perform more medical activities. Physician, who experiences revenue pressure, may try to increase his/her revenue by performing more medical activities and give less importance to quality. Resulting inadequate treatments and wrong diagnosed patients would have negative effects on health quality. On the other hand, physicians who do not have any revenue concerns may give the quality of healthcare absolute priority, undermining the productivity. This tendency may result in hospital crowding and high crowding pressures on physicians. Such conflicting pressures are included in model to investigate the impacts of P4P on health outputs in public hospitals. Results obtained concur with our dynamic hypotheses and agree with some of the general behaviors recently observed in Turkish healthcare.


Researcher:
Tuğrul MEKER
Contact:
ybarlas@boun.edu.tr

Deciding on Parameter Values of Anchor and Adjust Heuristic in Stock Management
Anchor-and-adjust is a widely used heuristic in stock management because, it is a fine representation of human decision making process in managing a stock. This thesis focuses on two issues related to the usage of the anchor-and-adjust heuristic in stock management: One is the selection of the decision making parameter values and the other is the determination of the desired supply line values for multi-supplier systems. Weight of Supply Line and Stock Adjustment Time are two of the decision parameters of anchor-and-adjust heuristic. We seek a rule of thumb for assigning good values to them. We first introduce a new parameter that we call Relative Aggressiveness, which together with Weight of Supply Line; determine the nature of the stock behavior. Relative Aggressiveness is Acquisition Delay Time (delay duration) divided by Stock Adjustment Time. We propose 4 as a sufficiently good and applicable value for Relative Aggressiveness. In other words, we suggest taking Stock Adjustment Time as a quarter of Acquisition Delay Time. We also give optimal values of Weight of Supply Line in a table for different delay orders and Relative Aggressiveness values. Desired Supply Line is the product of expected acquisition lag and desired acquisition rate. This calculation ensures that supply line would produce the desired acquisition rate given that it is at this desired level. A wrongly calculated Desired Supply Line value leads to a steady-state error preventing stock approach its goal. Therefore, correct calculation of Desired Supply Line values is crucial. Desired acquisition rate is equal to the expected loss flow in a single-supplier system. However, it is not easy to decide on the desired acquisition rates for a multi-supplier system. We give a general formula for the calculation of Desired Supply Line values based on the supplier utilization priorities and supplier production/shipment capacities.


Researcher:
Ahmet MUTALLİP
Contact:
hakan.yasarcan@boun.edu.tr

Control Heuristics for the Soft Landing Problem
In this thesis, we first construct a basic model representing the soft landing problem. The aim of the modeling effort is to transparently represent the process of landing a spacecraft on the surface of a celestial body. The process of landing is an interesting problem because there are two main contradictory performance criteria to be met simultaneously; the landing duration should be as short as possible, but at the same time crashing the spacecraft to the surface should be avoided. If the only criterion was to prevent crashing the spacecraft, that would not be difficult to achieve by slowing down the landing process. However, long landing duration necessitates extensive use of fuel, which should also be avoided. As a summary, the main goal in the soft landing problem is to land the spacecraft as gently and as fast as possible. Many real life complexities such as delays caused by actuators and measurement processes are not represented in the model. Even under the simplifying model assumptions, the main goal of the soft landing problem still remains a challenging one because the two state variables Height and Velocity can only be indirectly controlled. In this work, we studied four different control heuristics for the soft landing problem. The first heuristic is adapted from the mass-spring-damper model using the similarity of the equations of the soft landing model developed to the equations of the mass-spring-damper model; both models can be reduced to a second order linear differential equation. The second one is a bang-bang heuristic that first allows the spacecraft to fall freely, but after a critical point is reached, it uses the reverse force thruster at its maximum power until the touchdown. Bang-bang heuristic minimizes the time needed to land. However, it is not as robust as the mass-spring-damper heuristic in the sense that it is more sensitive to deviations from model assumptions. The third heuristic is a combination of the bang-bang and mass-spring-damper heuristics. This new heuristic also borrows the concept Weight of Supply Line from System Dynamics literature. This new heuristic reconciles the two heuristics reducing their respective problematic behaviors. The last heuristic is the terminal guidance heuristic. Although, the formulations and the dynamic behavior of the terminal guidance heuristic are different than the new heuristic, its landing performance is very similar to the performance of the new heuristic. The mass-spring-damper, bang-bang, new, and terminal guidance heuristics are compared in terms of their performances in the presence of an error in the parameter estimates, an error in the height readings, and an overlooked factor such as a delay in changing the level of the force created by the reverse force thruster, which is known as actuator delay. Terminal guidance heuristic and new heuristic lie in between mass-spring-damper heuristic and bang-bang heuristic in the sense that they require a more reasonable time to land as compared to the mass-spring-damper heuristic and they are not as sensitive as the bang-bang heuristic to the deviations from the original model. Finally, constant mass assumption is relaxed to observe a potential change in the behaviors generated by the heuristics, including the deviations due to errors and actuator delay. This relaxation also enables a comparison for the fuel consumption values of the heuristics.


Researcher:
Togay TANYOLAÇ
Contact:
hakan.yasarcan@boun.edu.tr

Systemic Complexity, Learning, and the Role of Interactive Simulators
Interactive simulators are often used to help and accelerate learning in and about such complex systems. But there is little evidence in the literature about to what extent conceptual and ‘deep’ learning can be acquired by playing with interactive simulators. The main purpose of this research is to assess the extent of learning, by testing the ‘transfer’ of learning from one gaming situation to another. Transfer of learning is only possible by a conceptual understanding of the system. Thus, the learner should not only be able to show good performance in a given decision game, but should also learn the underlying structure. As an integral part of the research, proper measures of ‘systemic complexity’ will be defined. Several simulators with different contexts and levels of complexity will be built and experiments will be carried out to measure the subjects’ transfer of learning from one simulator to another.


Researcher:
Onur ÖZGÜN
Contact:
ybarlas@boun.edu.tr

Dynamic Modeling of Peritoneal Dialysis and Its Implementations in Children with Chronic Renal Failure
This study has been conducted in order to shed light on the effects of Peritoneal Dialysis (PD) treatment on the dynamic interactions of nutritional intake decisions and the growth and development of children with Chronic Kidney Failure. In the context of this study, the interrelationship of the substances which play a major role in developmental and vital indicators of the child-patients and their relationships with PD has been analyzed by using System Dynamics methodology and Stella Modeling/simulation software. At the end of the study, based on the constructed model, an interactive simulation game which represents the relationship between the diet and the ratios of accumulated toxic and beneficial materials in the body has been designed. With the help of such a game, doctors, patients and their families will be able to search for diet recipes suitable to patients’ health status.


Researchers:
Elvan GÖKALP
Gülsevi BAŞAR
Duygu TEKİN
Contact:
ybarlas@boun.edu.tr

Modeling the Dynamics of Thyroid Hormones and Related Disorders
The objective of this work is to construct a simulation model to portray the dynamics of certain common thyroid disorders and to provide a platform for scenario analysis to support medical education, training and research, without risking patients’ health. Firstly, Graves’ disease, the most common source of hyperthyroidism, is addressed. Goiter formation, effect of iodine availability on the severity of the disease, and increased T3/T4 –a commonly used diagnostic measure in hyperthyroidism– are all well captured by the model. Secondly, iodine deficiency, one prevailing cause of hypothyroidism, is discussed. The model was able to depict all the characteristic changes including the goiter formation and increase in T3/T4. Thirdly, the transient inhibitory effect of excessive iodine intake on thyroid gland is discussed. The model is able to demonstrate the enlargement in thyroid volume and the mild decline in thyroid hormones. Lastly, a condition called subacute thyroiditis, a common disorder in which thyroid gland is exposed to inflammation, is analysed. The typical triphasic clinical course of subacute thyroiditis is well represented by the model. In conclusion, with respect to both qualitative and quantitative information in literature, and interviews with the medical doctors, the model exhibits an acceptable degree of validity and is able to cover a wide range of thyroid-related disorders.


Researcher:
Oylum ŞEKER
Contact:
oylum.seker@boun.edu.tr

A Dynamic Simulation Model for Long Term Bone Mass Homeostasis and Osteoporosis
Osteoporosis is a skeletal disorder related to low bone mass and increased risk of fracture. Although it is seen in both sexes, the disease is more prevalent among women after menopause. The goal of this thesis is to construct a dynamic simulation model that can realistically reproduce long-term behaviour of postmenopausal bone loss. For this purpose, a system dynamics model is built which focuses on bone’s mechanical properties and its interactions with calcium homeostasis system in blood. Comparisons with available data indicate that the model realistically reproduces the behaviour of bone loss for both menopausal and nonmenopausal causes in women. Experiments with the model demonstrate that keeping peak strains above disuse threshold is essential in bone health. Both medical and non-medical interventions work for treating bone loss after menopause, but drug therapies are most influential in treating osteoporosis. For avoiding non-menopausal losses, simulation experiments show that calcium supplementation is essential in pre- and post-menopausal years.


Researcher:
Canan HERDEM
Contact:
ybarlas@boun.edu.tr

Random Number Generation Using Chaotic Dynamical Maps
In this study, we propose some chaotic functions to generate pseudorandom numbers (PRNs), using the unpredictability property of dynamical chaotic maps. We suggest five different random number generators (RNGs) that are derived from three different chaotic maps: tent map, logistic map, and family of connecting maps. The uniformity and independence of the numbers generated through the five suggested RNGs are checked in three steps. Firstly, the histograms and serial plots are visually checked. Secondly, chi-square and Kolmogorov-Smirnov tests are applied to statistically test the uniformity of the generated numbers. Finally, runs tests and autocorrelation test are applied in order to check the independence of the numbers. The same tests are applied to compare the five suggested chaotic generators with some well-known conventionally used LCGs. It is concluded that the suggested generators perform nearly as well as LCGs and can lead to an alternative way of generating random numbers. More detailed mathematical, statistical, and numerical properties of the suggested generators constitute useful further research topics.


Researcher:
Ali Rıza ÖZÖREN
Contact:
ybarlas@boun.edu.tr

Modeling the Dynamics of Academic Publications and Citations
In this study, a dynamic simulation model is developed for analyzing the change in publication practices of researchers towards improving the performance measures used. Reputation of the faculty, skill level, total time devoted to research activities, acceptance fraction of the papers by the journals, publication and citation pressure on the researchers are the basic variables in the model. The model is constructed and calibrated using Boğaziçi University Engineering Faculty data. Forcing researchers to publish in high numbers results in numerous spurious publications with lower citations. Upgrading standards by introducing high benchmarks makes researchers increase the time that they devote to research activities. Allowing researchers spend more time on research activities by easing their loads is found to be an effective managerial policy. Encouraging mostly the high quality research is found to be a good policy, by generating more high quality publications compared to low quality ones, hence increased citations.


Researcher:
Nisa ÖNSEL
Contact:
nisa.guler@boun.edu.tr

A Methodology for Statistical Sensitivity Analysis of System Dynamics Models
The effects of possible changes in parameter values to the system behaviour should be questioned in order to develop an insight to the structure of the system. In this study, the usage of behaviour measures is proposed for sensitivity analysis of system dynamics models. The sensitivity of system behaviour is calculated by means of regression and ANOVA of scatter plots. The regression is found to be successful for finding the important parameters of the model for different behaviour measures. Also, analysis with different behaviour modes indicate that each behavior mode is sensitive to different set of model parameters. Therefore to make a consistent sensitivty analysis, it is necessary to separate different behaviour modes from each other. All results of regression methodology are confirmed with the ANOVA of scatter plots.


Researcher:
Mustafa HEKİMOĞLU
Contact:
ybarlas@boun.edu.tr

The Dynamics of Publication and Citation Networks in Academia
Three aspects of scientific publication and citation dynamics are analyzed in this study. Firstly, it is shown that average publication statistics and tendencies are different in different fields. Being the second part of the study, a system dynamics model describes how academicians behave under publication and citation pressure. Lastly, an agent based model is constructed showing that young scholars, who try to enter into fields which are filled by senior authors, have difficulty in receiving enough citation, compared to the ones who are trying to enter into fields with junior academicans. This suggests that comparing research performances of young researchers by just publication and citation numbers may not be valid and fair.


Researcher:
Burak ESKİCİ
Contact:
ybarlas@boun.edu.tr

Simulation Modeling of Body Weight Dynamics and Web Game Development
In this study, a simulation model is constructed based on Hall’s body weight simulation model. In the first part, some simplifications and modifications are made on the original model, while maintaining validity. In the second part, a software (Mashap) is developed that automatically generates and runs a web based simulation game for any system dynamics model. Finally, the game generator is used to develop a web based game for the modified body weight dynamics model. Game generator will enable system dynamics researchers to build web based gaming environments for their own models.


Researcher:
Mert NUHOĞLU
Contact:
ybarlas@boun.edu.tr

A Simulation Model for Blood Cholesterol Dynamics and Related Disorders
In this study, system dynamics method, which has a systemic view that other cholesterol studies lack, is employed in modeling and analyzing the dynamics of cholesterol metabolism. For both healthy and hypercholesterolemic subjects the model is able to generate realistic behaviors of different types of blood cholesterol, and body weight. It is observed that exercise is more effective than diet. In the case of hypercholesterolemic patients, the model effectively mimics the way the drugs work and shows how the patient can reach healthier cholesterol levels.


Researcher:
Emre M. DEMİREZEN
Contact:
ybarlas@boun.edu.tr