Přístupnostní navigace
E-application
Search Search Close
Master's Thesis
Author of thesis: Bc. Jan Píža
Acad. year: 2025/2026
Supervisor: Ing. Václav Miklas, Ph.D.
Reviewer: Mgr. Jaromír Sobotka, Ph.D.
This master’s thesis deals with the design of a virtual sensor for the estimation of Chemical Oxygen Demand (COD) in wastewater applications. Virtual sensors represent an economically and time-efficient alternative to the direct measurement of difficult-to-measure variables, enabling the estimation of monitored parameters and more efficient process control. The aim of the thesis was to evaluate current approaches to virtual sensors, collect data for the design of a virtual sensor, analyze relationships between measured variables, and subsequently develop a predictive model. The theoretical part summarizes approaches to virtual sensors, statistical methods, and machine learning techniques. The experimental part focused on the preparation of synthetic wastewater corresponding to a real dairy industry operation and on COD measurements together with selected input variables identified through a literature review, specifically turbidity, pH, ORP, and conductivity. Data analysis demonstrated a very strong correlation between COD and turbidity, while the remaining variables showed weak and practically less applicable relationships. Various models were tested for the development of the virtual sensor using the Python programming language and the PyCaret library. The best results on synthetic data were achieved by the Extra Trees Regressor model with coefficient of determination of R2 = 0,98 a mean absolute error of 163.7 mg/l. However, this model did not demonstrate sufficient extrapolation capability when validated on real wastewater data. From a practical point of view, linear (R2 = 0,97, MAE = 256,1 mg/l) and power-law (R2 = 0,97) models based solely on turbidity proved to be the most suitable. The thesis also includes a conceptual design of a virtual sensor based on continuous turbidity measurement and mathematical COD estimation in real time. The main benefit of the proposed sensor lies in continuous wastewater quality monitoring and early identification of abnormal operating conditions. A key factor affecting long-term reliable operation is regular maintenance and prevention of sensor fouling.
virtual sensor, chemical oxygen demand, wastewater, turbidity, dairy industry, Python, regression model
Date of defence
09.06.2026
Result of the defence
Defended (thesis was successfully defended)
Grading
A
Process of defence
Byla předvedena prezentace závěrečné práce a zodpovězeny otázky oponenta práce. Doplňující dotaz na provedená ověření v kontextu závěrečné práce, bylo zodpovězeno. Otázka na způsob měření turbidity, bylo zodpovězeno. Dotaz na rozsah měření v předvedené práci a související přenositelnost výstupů závěrečné práce na odlišný typ odpadních vod, bylo zodpovězeno. Dotaz na význam zanášení v kontextu práce a možnosti prevence zanášení, bylo zodpovězeno.
Language of thesis
Czech
Faculty
Fakulta strojního inženýrství
Department
Institute of Process Engineering
Study programme
Process Engineering (N-PRI-P)
Composition of Committee
prof. Dr. Ing. Marcus Reppich (předseda) doc. Ing. Martin Pavlas, Ph.D. (místopředseda) prof. Ing. Petr Stehlík, CSc., dr. h. c. (člen) prof. Ing. Zdeněk Jegla, Ph.D. (člen) Ing. Pavel Skryja, Ph.D. (člen) Ing. Pavel Lošák, Ph.D. (člen) Mgr. Ing. Marek Vondra, Ph.D. (člen)
Supervisor’s reportIng. Václav Miklas, Ph.D.
Grade proposed by supervisor: B
Reviewer’s reportMgr. Jaromír Sobotka, Ph.D.
Grade proposed by reviewer: B
Responsibility: Mgr. et Mgr. Hana Odstrčilová