Detection, Monitoring, and Simulation of Progressive Damage in Wind Turbine Blades

Mads Anker Fremmelev*

*Corresponding author for this work

Research output: Book/ReportPh.D. thesis

107 Downloads (Pure)

Abstract

The purpose of this PhD thesis is to conduct a feasibility study on structural health monitoring of large-scale wind turbine blades. Toward this goal, an extensive test campaign is performed, which at the time of writing is one of a kind in terms of blade size, variety of sensing systems, number of damage cases, and the intermediate repair of damages. A 52-meter commercial wind turbine blade is used, in which four different artificial damages are introduced, all being representative of damage that could be expected to occur in blades on operating turbines. Various sensing systems are used to collect data, using well-established sensing systems as well as a novel active vibration system specifically designed for this project.

The PhD project is motivated by a desire to design a system that enables structural health monitoring of blades on operating wind turbines. Such a system would reduce the need for scheduled manual inspections of wind turbine blades and decrease the risk of blade failure, both of which would contribute to reducing the cost of wind energy.

The blade test, which forms the foundation of this project through the real-life data acquired, is presented in Paper A. The following sensor types and sensing systems are used: electrical resistance strain gauges, fiber Bragg gratings, digital image correlation, acoustic emission, accelerometers, an active vibration monitoring system, and a guided waves system. For some of the listed sensing systems, their opportunities and limitations for detection of initiation and progression of damages in wind turbine blades are investigated. It is found that most of the investigated sensing systems are useful for either local or wide-range detection and monitoring of damages in wind turbine blades. The utilized local monitoring systems are found to have a useful detection range of up to 0.5 and 2 meters for acoustic emission sensors and strain gauges, respectively, given their specific placement and the damage cases investigated. A computational model is established to predict the strain behavior near damages, and good agreement is found between simulations and strain measurements from the blade test. Regarding global damage detection in wind turbine blades, it is found that the global eigenfrequencies and mode shapes of the blade are not sensitive to smaller damages. Lastly, an active vibration-based monitoring system specifically designed for the blade test shows opportunities in detecting initiation and progression of small damages within distances of at least ten meters.

Paper B explores the potential for using an active vibration monitoring system for blade structural health monitoring. From the recorded time series data, given a sinusoidal chirp active vibration input, the frequency content of acceleration measurements at different locations is calculated. Based on the frequency content of the acceleration measurements, features are extracted and used to calculate a damage index. From observations in the healthy state of the blade, a threshold value for the healthy state is defined, and any damage indices exceeding this value are classified as damaged observations. The classification accuracy is found to be very satisfactory, with monotonic progression of the damage index in line with damage progression being shown for multiple sensors. Thus, the active vibration monitoring system is shown to enable detection and monitoring of damages in wind turbine blades within large areas.

Lastly, Paper C investigates the use of different signal processing methods for extraction of damage-sensitive features. Based on an active vibration monitoring system, time-domain and time-frequency-domain methods are used to extract features from accelerometer data. For this purpose, sinusoidal signals with single frequencies and Gaussian envelopes are used, as opposed to the sinusoidal chirp signal used in Paper B. Using the extracted features, outlier analysis is performed to detect the initiation and progression of damage in the tested wind turbine blade. The performance in detection of initiation and progression of damage through use of the different features is compared. Generally, it is shown that high performance can be obtained through use of the active vibration monitoring system and different signal processing algorithms. Shortcomings in the use of single frequencies are discussed, and future best practices for practical application are laid out.

Collectively, Papers A, B, and C contribute to the field of wind turbine blade structural health monitoring by conducting an experimental investigation on damage initiation and progression in large-scale wind turbine blades; using various sensing systems to detect initiation and progression of damages; presenting the practical design and implementation of an active vibration monitoring system; and investigating signal processing methods for extraction of damage-sensitive features, enabling classification of the health state of blades.
Original languageEnglish
Place of PublicationAalborg, Denmark
PublisherDTU Wind and Energy Systems
Number of pages146
DOIs
Publication statusPublished - 2022

Fingerprint

Dive into the research topics of 'Detection, Monitoring, and Simulation of Progressive Damage in Wind Turbine Blades'. Together they form a unique fingerprint.

Cite this