Unbalance evaluation of a scaled wind turbine under different rotational regimes via detrended fluctuation analysis of vibration signals combined with pattern recognition techniques
Abstract This work aims to propose a different approach to evaluate the operating conditions of a scaled wind turbine through vibration analysis. The turbine blades were built based on the NREL S809 profile and a 40-cm diameter, while the design blade tip speed ratio ( λ ) is equal to 7. Masses weighing 0.5, 1.0, and 1.5 g were added to the tip of one or two blades in a varying sequence with the intent of simulating potential problems and producing several scenarios from simple imbalances to severe rotor vibration levels to be compared to the control condition where the three blades and the system were balanced. The signals were processed and classified by a combination of detrended fluctuation analysis with Karhunen-Loeve Transform, Gaussian discriminator, and Artificial Neural Network, which are pattern recognition techniques with supervised learning. Good results were achieved by employing the above cited recognition techniques as more than 95% of normal and imbalanced cases were correctly classified. In a general way, it was also possible to identify different levels of blade imbalance, thus proving that the present approach may be an excellent predictive maintenance tool for vibration monitoring of wind turbines.
Paper
Full text
Unbalance evaluation of a scaled wind turbine under different rotational regimes via detrended fluctuation analysis of vibration signals combined with pattern recognition techniques
Semantic Scholar · Engineering · 2019
Abstract
Abstract This work aims to propose a different approach to evaluate the operating conditions of a scaled wind turbine through vibration analysis. The turbine blades were built based on the NREL S809 profile and a 40-cm diameter, while the design blade tip speed ratio ( λ ) is equal to 7. Masses weighing 0.5, 1.0, and 1.5 g were added to the tip of one or two blades in a varying sequence with the intent of simulating potential problems and producing several scenarios from simple imbalances to severe rotor vibration levels to be compared to the control condition where the three blades and the system were balanced. The signals were processed and classified by a combination of detrended fluctuation analysis with Karhunen-Loeve Transform, Gaussian discriminator, and Artificial Neural Network, which are pattern recognition techniques with supervised learning. Good results were achieved by employing the above cited recognition techniques as more than 95% of normal and imbalanced cases were correctly classified. In a general way, it was also possible to identify different levels of blade imbalance, thus proving that the present approach may be an excellent predictive maintenance tool for vibration monitoring of wind turbines.