Symmetry-Based Fault Detection and Diagnosis for Hydraulic Wind Turbine Pitch Systems
Authors
Jensen, Mads Norman Kaae ; Petersen, Benedikte ; Kjeldsen, Thomas
Term
4. term
Education
Publication year
2025
Submitted on
2025-05-28
Pages
100
Abstract
Internal leakage is one of the most common faults in hydraulic pitch systems that set the angle of wind turbine blades. This thesis presents a new symmetry-based method: compare the behavior of signals from the three blades and detect faults from their relative differences. To study the method, a dynamic model of a pitch system was built and validated on a physical test setup, then extended to all three blades. OpenFAST was used to mimic realistic turbine dynamics. A pitch-angle control strategy was designed and implemented to enable simulation and testing under operating conditions. By converting signals to the blade-angle domain (aligning data by blade angle rather than time), clear patterns linked to internal leakage, friction, and external leakage were found. Internal leakage consistently causes both chamber pressures in the affected cylinder to drop compared with the two healthy cylinders. The same tendencies were observed on the physical setup. Based on these deviations, a residual-based detection and diagnosis algorithm was developed. It combines statistics (a threshold on residuals, i.e., differences between expected and measured behavior) with simple logic that diagnoses the fault from the pattern. In simulation, the algorithm diagnosed leaks as low as 0.06 L/min at rated wind speed and 0.3 L/min above rated speed. Above rated speed, the leakage flow was estimated with a maximum error of 0.3 L/min. Experiments confirmed that the method can detect internal leakage in hardware. The approach is simple and effective and, with further research, could be implemented on operating wind turbines.
Intern lækage er en af de mest udbredte fejl i hydrauliske pitchsystemer, som indstiller rotorbladenes vinkel. Denne afhandling præsenterer en ny metode baseret på symmetrien i den trebladede rotor: sammenlign signalerne fra de tre blade og find fejl ud fra deres indbyrdes afvigelser. For at undersøge metoden er der udviklet en dynamisk model af et pitchsystem, som er valideret på en fysisk testopstilling og udvidet til alle tre blade. OpenFAST er anvendt til at efterligne realistisk turbindynamik. En styringsstrategi for bladvinklen er designet og implementeret, så systemet kan simuleres og testes under drift. Ved at omdanne signaler til bladvinkeldomænet (data afstemmes efter bladvinkel i stedet for tid) fremtræder tydelige mønstre for intern lækage, friktion og ekstern lækage. Intern lækage medfører, at begge kammertryk i den berørte cylinder falder i forhold til de to raske cylindre. De samme tendenser blev observeret på den fysiske opstilling. På baggrund af disse afvigelser er der udviklet en residualbaseret detektions- og diagnosealgoritme. Den kombinerer statistik (en tærskel på residualer, dvs. forskelle mellem forventet og målt adfærd) med simpel logik, der diagnosticerer fejlen ud fra mønstret. I simulation blev lækager helt ned til 0,06 L/min ved nominel vindhastighed og 0,3 L/min over nominel hastighed diagnosticeret. Over nominel hastighed blev lækagemængden estimeret med en maksimal fejl på 0,3 L/min. Eksperimenter bekræftede, at metoden kan detektere intern lækage i hardware. Tilgangen er enkel og effektiv og kan med yderligere forskning implementeres i idriftsatte vindmøller.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords