In this thesis, we address the development of a digital twin of a hydraulic system and the application of machine learning for diagnosing hydraulic cylinder seal damage. We focus on the need for more effective condition monitoring of hydraulic components and on the development of methods that can support diagnostics and predictive maintenance. Particular attention is given to hydraulic cylinder seals, as their wear affects system operation and may cause internal leakage and changes in measured quantities. As part of the work, we used a hydraulic test rig on which measurements were performed under different operating conditions and levels of seal damage. We monitored piston-side and rod-side pressure, hydraulic fluid temperature, piston rod displacement, and flow rate. Based on the real system, we developed a simulation model in Automation Studio. The measured data were processed in Jupyter Notebook and used to develop a regression-based machine learning model for relative assessment of seal condition. By combining real measurements, simulation modelling, and data analysis, we established a basis for developing a hydraulic-system digital twin and for further development of methods for condition monitoring, diagnostics, and predictive maintenance of hydraulic components.
|
