Diagnosis of Low-Speed Bearings via Vibration-Based Entropy Indicators and Acoustic Emissions
- DIRECTORS: Francesc Pozo, Yolanda Vidal and Urko Leturiondo
- UNIVERSITY: Universitat Politècnica de Catalunya
Wind energy is one of the main renewable energies to replace fossil fuels in the generation of electricity worldwide. To enhance and accelerate its implementation at a large scale, it is vital to reduce the costs associated with maintenance. As component breakages force the turbine to stop for long repair times, the wind industry must switch from the old-fashioned preventive or corrective maintenance to condition-based maintenance (also called predictive maintenance). The condition-based maintenance of pitch bearings is especially challenging, as the operating conditions include high mechanical stress and low rotational speed. Since these operating conditions impact negatively on the results of the standard methods and techniques applied in current condition-based monitoring systems, the condition-based maintenance of pitch bearings is still a challenge.
Therefore, this thesis is focused on the research of novel methods and techniques that obtain reliable information on the state of pitch bearings for condition-based maintenance. Initially, the acknowledgment of the state of the art is performed to recognize the methods and signals. This step endorses the decision to analyze the vibration signals and acoustic emissions throughout this thesis. Due to the particular operating conditions of pitch bearings, this research states the need to create datasets to replicate the particular operating conditions in a controlled laboratory experiment. As a result, a dataset based on vibrations, and a second dataset based on acoustic emissions are generated.
The vibration dataset allows the validation of a novel algorithm for the low-speed bearing diagnosis, which is based on the concept of entropy by the definition of Shannon and Rényi. In comparison to the classical methods found in the literature, the diagnosis of low-speed bearings based on entropy-based indicators can extract more reliable information. Moreover, the research of the combination of several indicators to improve the diagnosis reveals that the entropy-based indicators can extract more information than regular indicators used in academia.
The dataset of acoustic emissions from low-speed bearings helps to contribute to the development of methods for diagnosis. In this research, the analysis of the energy from the signals reveals a dependency on the intensity and the presence of damage. In addition, a relation between the waveform of the analyzed energy and the existence of damage is emphasized.