Tez Türü: Doktora
Tezin Yürütüldüğü Kurum: Ankara Yıldırım Beyazıt Üniversitesi, Mühendislik Ve Doğa Bilimleri Fakültesi, Elektrik-Elektronik Mühendisliği, Türkiye
Tez Danışmanı: İlyas Çankaya
Tezin Onay Tarihi: 2025
Tezin Dili: İngilizce
Özet:
Glucose, one of the most essential carbohydrates in nature, plays a central role in the energy metabolism of living organisms. Therefore, its monitoring holds considerable significance in both healthcare and industrial applications. In this thesis, a sliding window algorithm has been developed to determine the frequency point where the separation is most apparent at increasing glucose concentrations to be applicable in biotechnology and medicine. In the initial phase, measurements were conducted using a vector network analyzer on various glucose-containing liquids, including Coke, Diet Coke, Coke Zero, soda, sugar-free soda, and different concentrations of glucose solutions. When the reflection coefficient results were analyzed, it revealed distinctive responses corresponding to increasing glucose concentrations. Subsequently, to enhance measurement sensitivity, key experimental parameters such as container material (PTFE, PLA, and ABS), wall thickness (0.6 mm and 1.0 mm), and sample volume (9, 45, and 90 cm³) were systematically optimized. The results indicated that the 1.0 mm wall thickness PTFE container offered the most favorable sampling condition, and larger sample volumes generally contributed to improved measurement accuracy. In the next stage, to more realistically simulate biotechnological environments, nitrogen-containing compounds such as ammonium sulfate and yeast extract were also evaluated using a similar methodology. A sliding window algorithm was proposed and implemented to identify frequency-specific regions that differentiate glucose and nitrogen sources. Finally, glucose solutions at the gram and milligram scales were analyzed using machine learning models—Decision Tree, Random Forest, and Support Vector Machines—with Random Forest achieving the highest predictive accuracy