Structural health monitoring of wind turbines: method and application to a HAWT

Abstract

Structural health monitoring in the context of a Micon 65/13 horizontal axis wind turbine was described in this paper as a process in statistical pattern recognition. Simulation data from a calibrated model with less than 8% error in the fi rst 14 natural frequencies of vibration was used to study the operational response under various wind states as well as the effects of three types of damage in the blade, low speed shaft and yaw joint. It was shown that vertical wind shear and turbulent winds lead to different modal contributions in the operational response of the turbine suggesting that the sensitivity of operational data to damage depends on the wind loads. It is also shown that there is less than a 4% change in the wind turbine natural frequencies given a 25% reduction in the stiffness at the root of one blade. The modal assurance criterion was used to analyse the corresponding changes in modal defl ections, and this criterion exhibited nearly orthogonal changes because of the three damage scenarios suggesting that the modal defl ection determines which damage is observable at a given frequency for a given wind state. The` modal contribution is calculated as a damage feature, which changes as much as 100% for 50% reductions in blade root stiffness, but only the blade damage is detected using this feature. Operational data was used to study variations in the forced blade response to determine the likelihood that small levels of damage can be detected amidst variations in wind speed across the rotor plane. The standard deviation in measured data was shown to be smallest for the span and edge-wise measurements at 1P due to gravity, which provides the dominant forcing function at this frequency. A 3% change in the response in the span and edge-wise directions because of damage is required to detect a change of three standard deviations in contrast to the 90% change in fl ap direction response that is required to detect a similar change because of damage. The dynamic displacement in the span direction is then used to extract a damage feature from the simulation data that provides the ability to both locate and quantify the reduction in stiffness in the blade root.

Effective Methods for Structural Health Monitoring of Critical Zones of Scalable Wind Turbine Blades

“…The monitoring of WTB made from GFRP have been reviewed in [9], proposing either statistical pattern recognition using simulation or experimental data, or in [19] considering mechanical property testing and full-scale testing as well as nondestructive testing methods. In order to calculate the lifetime of GFRP structures used in WTB construction, it is essential to have access to information about fatigue resistance of the material.…”

Damage detection on mesosurfaces using distributed sensor network and spectral diffusion maps

“…3(f) shows the cartesian coordinates of the nine virtual strain gauges. The simulated damage is a loss of stiffness arising from delamination, a damage mode commonly studied in wind turbine blade literature [3]. It is modeled as a change in the stiffness of laminate layer 2.…”

Value-Centric Approaches to Design, Operations, and Maintenance of Wind Turbines

“…On average, scheduled maintenance is performed about twice per year on installed turbines and results in an average downtime of 24 hours per turbine, per inspection (Adams et al, 2011). In contrast, unscheduled maintenance is 500% more costly on average, and results in an average downtime of 130 hours per turbine per year for European turbines (Adams et al, 2011).…”