Research Papers

Wind Turbine Blade Damage Detection Using Supervised Machine Learning Algorithms

[+] Author and Article Information
Taylor Regan, Christopher Beale

Department of Mechanical Engineering,
University of Massachusetts Lowell,
Lowell, MA 01854

Murat Inalpolat

Department of Mechanical Engineering,
University of Massachusetts Lowell,
1 University Avenue,
Lowell, MA 01854
e-mail: Murat_Inalpolat@uml.edu

1Corresponding author.

Contributed by the Technical Committee on Vibration and Sound of ASME for publication in the JOURNAL OF VIBRATION AND ACOUSTICS. Manuscript received June 10, 2016; final manuscript received May 26, 2017; published online August 2, 2017. Assoc. Editor: Patrick S. Keogh.

J. Vib. Acoust 139(6), 061010 (Aug 02, 2017) (14 pages) Paper No: VIB-16-1291; doi: 10.1115/1.4036951 History: Received June 10, 2016; Revised May 26, 2017

Wind turbine blades undergo high operational loads, experience variable environmental conditions, and are susceptible to failure due to defects, fatigue, and weather-induced damage. These large-scale composite structures are fundamentally enclosed acoustic cavities and currently have limited, if any, structural health monitoring (SHM) in place. A novel acoustics-based structural sensing and health monitoring technique is developed, requiring efficient algorithms for operational damage detection of cavity structures. This paper describes the selection of a set of statistical features for acoustics-based damage detection of enclosed cavities, such as wind turbine blades, as well as a systematic approach used in the identification of competent machine learning algorithms. Logistic regression (LR) and support vector machine (SVM) methods are identified and used with optimal feature selection for decision-making via binary classification algorithms. A laboratory-scale wind turbine with hollow composite blades was built for damage detection studies. This test rig allows for testing of stationary or rotating blades, of which time and frequency domain information can be collected to establish baseline characteristics. The test rig can then be used to observe any deviations from the baseline characteristics. An external microphone attached to the tower will be utilized to monitor blade health while blades are internally ensonified by wireless speakers. An initial test campaign with healthy and damaged blade specimens is carried out to arrive at several conclusions on the detectability and feature extraction capabilities required for damage detection.

Copyright © 2017 by ASME
Your Session has timed out. Please sign back in to continue.



Grahic Jump Location
Fig. 1

Schematic of the active damage detection

Grahic Jump Location
Fig. 2

Illustration of the feature distinguishability metric

Grahic Jump Location
Fig. 3

General process overview for the supervised learning algorithms

Grahic Jump Location
Fig. 4

The sigmoid hypothesis function, hθ(x)=g(z)

Grahic Jump Location
Fig. 5

Logistic regression cost curves for y = 1 and y = 0

Grahic Jump Location
Fig. 6

Multiple hyperplanes fitting sample dataset

Grahic Jump Location
Fig. 7

Illustration of the steps in hyperplane formulation

Grahic Jump Location
Fig. 8

Ill-fitting and optimal hyperplane comparison

Grahic Jump Location
Fig. 9

Cost functions for (a) y = 1 and (b) y = 0

Grahic Jump Location
Fig. 10

(a) Solid model of the subscale turbine and (b) completed subscale turbine prototype

Grahic Jump Location
Fig. 11

Schematic of blade 1 damage locations

Grahic Jump Location
Fig. 12

Instances of accuracies greater than 98%

Grahic Jump Location
Fig. 13

Combined damage case testing accuracies

Grahic Jump Location
Fig. 14

Machine learning accuracies for rotating tests

Grahic Jump Location
Fig. 15

Comparison of the peak amplitude FFT and mean frequency features for a stationary multi-mid excitation test

Grahic Jump Location
Fig. 16

Comparison of the RMS and kurtosis features for a stationary multi-high excitation test

Grahic Jump Location
Fig. 17

The two-feature plot for all test cases from a stationary multi-mid excitation test

Grahic Jump Location
Fig. 18

A side-by-side comparison of a feature pair with and without the first level of damage included

Grahic Jump Location
Fig. 19

A side-by-side two-feature plot comparing the variability of healthy data clusters across test cases

Grahic Jump Location
Fig. 20

Distinguishability distribution for (a) case 1, (b) case 2, and (c) case 3

Grahic Jump Location
Fig. 21

Multi-high tip hole distinguishability results

Grahic Jump Location
Fig. 22

Fisher's ratio for all tests and excitation types




Some tools below are only available to our subscribers or users with an online account.

Related Content

Customize your page view by dragging and repositioning the boxes below.

Related Journal Articles
Related eBook Content
Topic Collections

Sorry! You do not have access to this content. For assistance or to subscribe, please contact us:

  • TELEPHONE: 1-800-843-2763 (Toll-free in the USA)
  • EMAIL: asmedigitalcollection@asme.org
Sign In