Detection of hydraulic cylinder leakage

Lukjanec, Dmitrij ; Vietor, Arne ; Balint, Zoltan

4. semester

Sustainable Energy Engineering, Master

2016

2016-06-08

137

Dette speciale undersøger, hvordan man kan opdage og diagnosticere hydrauliske lækager i en kran-testopstilling ved hjælp af metoder til fejldetektion og -diagnose (FDD). To tilgange blev udviklet og sammenlignet: et udvidet Kalman-filter (EKF), som estimerer systemets indre tilstand ud fra støjfyldte målinger, og spektralanalyse, som søger efter fejlspor i signalernes frekvenser. Arbejdet omfattede laboratorieforsøg, verifikation af en matematisk model samt offline-simuleringer baseret på både model- og signaldatakilder. Studiet viser, at lækagens størrelse har stor betydning for, om den kan detekteres og isoleres, og at valget af en tærskelværdi for lækkoefficienter er afgørende. Den valgte tærskel gav korrekte resultater for lækager på niveau 4 og derover, men var ikke præcis for mindre lækager. Samlet set klarede EKF sig bedre end spektralanalyse. Spektralanalyse krævede omfattende for-analyse for at opstille pålidelige fejlindikatorer, mens EKF gav troværdige resultater for kunstige lækager ved eller over niveau 4 (0,1293–0,9853 l/min) ved 25°C. Fremtidigt arbejde bør gentænke EKF-strukturen ved at modellere lækagen som en systemtilstand og lade lækkoefficienter indgå som estimerede parametre.

This thesis investigates how to detect and diagnose hydraulic leaks in a crane test setup using fault detection and diagnosis (FDD) methods. Two approaches were developed and compared: an Extended Kalman Filter (EKF), which estimates the system’s internal state from noisy measurements, and spectral analysis, which looks for fault signatures in signal frequencies. The work combined laboratory experiments, verification of a mathematical model, and offline simulations using both model-based and signal-based data. The study shows that the size of the leak strongly affects whether it can be detected and isolated, and that choosing a threshold for leak coefficients is critical. The selected threshold correctly identified and isolated leaks at level 4 and higher, but was not accurate for smaller leaks. Overall, the EKF outperformed spectral analysis. Spectral analysis required extensive pre-analysis to set up reliable fault indicators, whereas the EKF provided trustworthy results for artificial leaks at or above level 4 (0.1293–0.9853 l/min) at 25°C. Future work should reconsider the EKF structure by explicitly modeling the leak as a system state and updating leak coefficients as estimated parameters.

[This abstract was generated with the help of AI]

fault ; detection ; diagnosis ; isolation ; hydraulics ; lagrangian ; pantograph ; spectral ; model based ; signal based

Download
View record in AAU Student Projects