Special Issue Article

Damage assessment of wind turbine blade under static loading test using acoustic emission

Journal of Intelligent Material Systems and Structures 0(0) 1–10 Ó The Author(s) 2013 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/1045389X13508329 jim.sagepub.com

Byeong-Hee Han1,2, Dong-Jin Yoon1, Yong-Hak Huh3 and Young-Shin Lee2

Abstract Acoustic emission is known as a powerful nondestructive tool to detect any further growth of preexisting cracks or to characterize failure mechanisms. Recently, this kind of technique, which is an in situ monitoring of integrity of materials or structures, becomes increasingly popular for monitoring the conditions of large structures such as a wind turbine blade. Therefore, it is required to find a symptom of damage progress before catastrophic failure through a continuous monitoring. In this study, acoustic emission technology was applied to assess the damage in the wind turbine blade during step-by-step static load test. In this static loading test, we have used a full-scale blade of 100 kW in capacity, and an attempt was made to apply a new source location method using a new algorithm with energy contour mapping concept. We also measured the deflection of blade tip by linear variable differential transformer (LVDT) and the strain of inner shear web in order to analyze the correlation between stress condition and damage identification. The results show that the acoustic emission activities give a good agreement with the stress distribution and damage location in the blade, especially in bonding edges around 1000–1500 mm far from the root. Finally, the applicability of new source location method was confirmed by comparison of the result of source location and experimental damage location. Keywords Wind turbine blade, acoustic emission, nondestructive evaluation, source location, composite materials

Introduction In order to harvest more energy with the considerations of high efficiency and cost-effectiveness, the size of the wind turbine blade has increased over the years. Usually, the size of blade is typically about 25 m for 750 kW and about 45 m for 2–3 MW in length. As wind turbine blades increase in size, there is an increasing need to monitor the integrity of the structures (Borum et al., 2006; Ciang et al., 2008). In addition, the technology for structural health monitoring (SHM) may allow the use of lighter blades that would provide higher performance with less conservative margins of safety (Schulz and Sundaresan, 2006). The wind turbine blades usually use several composite materials such as glass fiber-reinforced plastic (GFRP), polyvinyl chloride (PVC), and balsawood which ensure weight versus strength ratio. In addition, they have different thicknesses in the root and tip area of blade because of weight distribution and aerodynamic shape for efficiency of power generation. So, more appropriate nondestructive testing (NDT) method for evaluating the integrity of this kind of inhomogeneous materials or geometrically heterogeneous

structures is recently required. There are several structural damage factors for wind turbine blade such as incomplete permeation of resin in manufacturing process, adhesive missing in bonding process, and impact damages during transportation and installation (Sørensen et al., 2004). In addition, there will be many external factors to affect the integrity of wind turbine blade during normal operation. Typically, it is a delamination of composites by sudden wind gust, cracking by foreign objects impact, and natural disaster such as lightening, hail, and typhoon (Flemming and Troels, 2003; Ghoshal et al., 2000). 1

Center for Safety Measurements, Korea Research Institute of Standards and Science, Daejeon, Republic of Korea 2 Department of Mechanical Engineering, Chungnam National University, Daejeon, Republic of Korea 3 Center for Energy & Material Standard, Korea Research Institute of Standards and Science, Daejeon, Republic of Korea Corresponding author: Dong-Jin Yoon, Center for Safety Measurements, Korea Research Institute of Standards and Science, 267 Gajeong-ro, Yuseong-gu, Daejeon 305-340, Republic of Korea. Email: [email protected]

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

2

Journal of Intelligent Material Systems and Structures 0(0)

Acoustic emission technology (AET) has been used for the SHM of large structures such as civil structures such as a bridge, pressure vessel, and aerospace structures. Especially, comparing to other NDT technology, it is one of the most powerful techniques being able to detect damages and to identify damage location during operations. However, in case of the source location technique, there is some limitation in conventional AET because it strongly depends on wave speed in the corresponding material such as a wind turbine blade, especially inhomogeneous material or heterogeneous structures. That is, a new concept-based source location method is required in order to apply to complex composite structures. For instance, a new source location method in the cylindrical pressure vessel was developed, which is calculating arrival time delay according to each different propagation route in the cylindrical geometry (Yoon et al., 1990). The other new source location method was shown in smart active layer (SAL) sensor, that is, selfdesigned and embedded in composite structures. This source location algorithm also utilizes the arrival time delay and signal amplitude variation corresponding to the distance between sensor and damage source (Yoon et al., 2005). However, these two source location methods still depend on the wave velocity in the corresponding structures. In contrast to the other NDT methods which are usually carried out during overhaul period, acoustic emission (AE) technique is usually applied to structures under operation condition. In case of wind turbine blade, AE testing was applied for both static and fatigue loading tests in the laboratory scale experiments. In these studies, AE was used to detect damages in the blade inside. In addition, they compared real damage location with the AE source location results by using AE sensors patched on blade skin. Moreover, the results of AE activity and its analysis were compared with other NDT methods (Beattie, 1997; Jørgensen et al., 2004). In the signal processing point of view, a denoising technique was applied to AE testing to improve the accuracy of source location and signal identification. The application of a wavelet transform (WT) to identify AE sources was also carried out. The magnitudes in wavelet correspond to a particular group velocity and signal frequency for each mode, the ratios were able to distinguish different source types and exhibited only small changes with increasing propagation distance (Grosse et al., 2004; Hamstad et al., 2002). In cases of necessity of high accuracy of damage evaluation, usually, the number of sensors will be increased and consequently the number of data output to the signal processing system also increased. In order to reduce the number of data, they proposed a structural neural system (SNS) for SHM system. A highly distributed continuous sensor concept, which mimics the signal processing in the biological neural system, is adopted.

The continuous sensors for SNS are formed by individual lead zirconate titanate (PZT) sensors connected in a series (Kirikera et al., 2007, 2008). Another approach of source location for wind turbine blade which consists of heterogeneous material is tomographical method. In order to overcome these complex characteristics, the conventional AE source location algorithm can be combined with travel time tomography by using the AE events as acoustic point sources. In this work, a relocation of the current source positions has to be performed after each tomographic inversion, which results in an iterative procedure with alternating steps of source location and tomography (Schubert, 2004). Also, source location study using the neural network was done in order to overcome signal attenuation problem for sensor frequency (Caprino et al., 2011; Godin et al., 2004; Oliveiraa and Marquesb, 2008). All the above source location techniques basically detect elastic wave on the structures and then carry out source location using time arrival delay algorithm. Therefore, these algorithms are highly dependent on the propagation wave speed of the tested structures. Therefore, it is very difficult to locate AE source exactly because of its heterogeneous material property. Consequently, we need a new algorithm for accurate damage location and source characterization for wind turbine blade which consists of dissimilar composite materials. In this study, full-scale blade was tested in static loading condition in order to verify the performance of new developed source location method. This source location technique is based on the energy contour map which has a unique pattern corresponding to each blade (Yoon and Han, 2012). In addition, we have also tried to measure strain and deflection of blade during stepwise static loading scheme in order to compare the AE activity with stress history.

Source location using energy-based counter map Usually, the limitation of conventional AE source location method using arrival time difference is strongly revealed by the dependence of wave speed in the corresponding material of tested structures, especially inhomogeneous material or heterogeneous structures. In case of very thick composite blade, it is very hard to measure AE signal in conventional normal sensor frequency ranges because these kinds of dissimilar composite materials have high attenuation property and very thick geometry in skin thickness. Therefore, new proposed source location method should be less affected by the wave speed in these kinds of composite blades. That is, the measurement of energy distribution in the composite materials is better

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

Han et al.

3

than time arrival method in its reproducibility point of view. In addition, it will be better to install minimum number of sensors on the structures to be covered. In order to satisfy these conditions, we have developed a new source location algorithm which is based on the AE energy contour map (Yoon and Han, 2012). Figure 1 shows the input source generation process which indicates the position of sensor and input data point to be tested. In case of Figure 1, totally 4 sensors and 144 input points were used to calculate AE signal intensity and construct a contour map of blade to be tested. After conducting the several steps of signal processing, database contour map was completed for each sensor as shown in Figure 2. Usually, one AE sensor is needed to get one contour map for the interested area of blade. This contour map indicates a pattern of signal attenuation corresponding to the position of sensor and input source. The procedure of measurement for a new source location algorithm is as follows: First, the area to be monitored will be decided in the tested blade and then the optimal number of sensors on the blade which can cover this interested whole area of the blade will be decided. Next, the input source point should be chosen in the monitoring area to accomplish a contour map as shown in Figure 1. After these procedures, the input source on each point has to be applied using several kinds of impact sources such as pencil lead break, steel ball impact, and so on. In this case, 144 input point sources have to be used. Then, the AE signal corresponding to each input source from each sensor is measured. Finally, a 16 3 9 matrix database for each sensor will be obtained, which displays a contour map as shown in Figure 2. This database consisted of each intensity value acquired from this pre-set data point on the blade surface before installing the blade onto the tower and nacelle. Each contour line indicates same value of energy obtained from each corresponding sensor on the blade. That is, this line shows an absolute value calculated

from power spectrum density of received AE signal. Even though these results are correlated to characteristics of wave propagation such as attenuation and mode conversion, this contour line implies unique information such as a fingerprint which depends on the path of wave propagation for the corresponding materials or structures. In this case, input source point will be decided with consideration of the size and shape of structures, and the interval of points can be adjusted by optimal resolution of source location. Consequently, these setup conditions will improve the algorithm efficiency and the result of source location. Considering several kinds of damages, each different arbitrary input source was used to get the initial database. Finally, you have to find a solution by comparing certain value between each sensor output and preacquired database of the interested blade. If the energy value measured from external unknown damage sources is overlapped to the value of pre-acquired database, you can find unique solution for a corresponding damage location. That is, source location (x, y) can be calculated from matrix coordinate (m, n). In these procedures, some numerical analyses such as iteration method, data comparing tolerance can be applied to optimize source location exactly. For example, the number of data of matrix is directly affected by the accuracy and error of the source location results. Therefore, the signal processing which minimizes the error through data interpolation is needed. This interpolation process of data will decrease the timeconsuming for generation of a contour map. Finally, the wave speed or three-dimensional (3D) effects in this energy contour map–based algorithm is not considered since this new source location algorithm uses only energy-based wave propagation response in the interested structural material itself. However, the database for the contour map will be affected if the materials or geometries in the structures are changed. That is, the energy contour mapping is unique if there

Figure 1. Input source generation process in the energy mapping source location method.

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

4

Journal of Intelligent Material Systems and Structures 0(0)

Figure 2. Contour map for each sensor.

Figure 3. Schematic diagram of static loading test system with whiffletree. AE: acoustic emission.

is no geometrical or material change in the tested structures.

Experiments Full-scale blade and static loading test In order to assess the damages in the wind turbine blade, we used a full-scale blade of 100 kW in capacity. This blade was specially designed for experiments, which is 11,000 mm long in length and root part of 560 mm in diameter. They consist of two skin plates and one single shear web in the middle of blade. This blade skin and shear web are made of GFRP and balsawood, respectively. A static loading test system with whiffletree structures for continuous load distribution of blade was set up. For this loading system, especially designed whiffletrees were attached on the wind turbine blade at three point supporters, which are connected to a

hydraulic-actuator using a steel wire as shown in Figure 3. Loading force was controlled by hydraulicactuator and also measured by load cell that is installed between actuator and whiffletree. Before the main loading test, we have a pre-loading test for checking the condition of loading scheme. In this primary test stage, unexpected initial damage occurred in the edge part of blade due to the error of blade design. The debonding damages are located in the leading and trailing edge with visible size, and the cracks occurred at the lower value than at the expected load. Therefore, the scheme of loading has been changed up to 90% of maximum design load and it was controlled by stepwise loading condition. This loading step was applied to a series of 20%– 40%–60%–70%–80%–90%–60%–40% of maximum design load, and it has a holding time of 20 min at every step for observing damage propagation as shown in Figure 7. In addition, the displacement was

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

Han et al.

5

Figure 4. AE sensor location in the tested blade: (a) 3D view of wind turbine blade and (b) location of embedded acoustic emission sensor (each value indicates a distance from the root). AE: acoustic emission; 3D: three-dimensional.

measured at the blade tip, and this static loading test was done by stroke control.

Sensor installation and experimental setup AE measurement was done during step-by-step static load test to observe the activity of damage in the blade. First, the resonant frequency of AE sensor was decided from the experiment, which has a best performance under consideration of wave propagation characteristics such as high attenuation, heterogeneous materials, and so on, that is, experimental selection by consideration of geometry and attenuation characteristic in the tested blade materials. Finally, we considered two types of AE sensors which have 30 and 60 kHz as resonant frequencies. In this experiment, six R6I-type sensors (PAC, USA) which have 60 kHz resonant frequency ranges were used, and this integrated type sensor includes a pre-amplifier inside. These six AE sensors were embedded in the blade skin using same epoxy, and also the sensors were fixed by special jig for 24 h until epoxy hardening. The position of embedded sensors is shown in Figure 4. The location of the sensor was decided by consideration of expected damages area which has stress concentration in geometrically. That is, it has a distance of 500, 2000, and 4000 mm from the root, and it was located intentionally closed to both edge sides as shown in Figure 4(b). We have used two measurement systems of commercial AE equipment and self-composed high-speed digitizer. MicroSAMOS AE system (PAC) was used for AE data acquisition and

analysis, and also a eight-channel high-speed digitizer was used to acquire raw data signal for applying of new proposed source location algorithm. The analysis for new source location algorithm was done by self-coded LabVIEW and MATLAB programs. A series of pre-database acquisition procedure was carried out before installation of blade for testing. In order to compose the contour map of new blade to be tested, three types of input sources, commercial Equotip impact tester type-C/D/G, which are commercial ball-drop-type impact sources, were used for elastic wave source generation. This Equo-tip apparatus is based on a spring-loaded impact body, which is propelled against a test piece. This tester is usually used for measuring the hardness of materials. We used the impact energy as AE sources, which have several different ranges in their capacity. The quantity of energy is C \ D \ G in the order of their capacity. In addition, there are also 18 strain gages that were attached on shear web in order to observe the strain behavior of blade as shown in Figure 5. The shear web is located vertically inside the blade skin. The electric strain gage was attached close to the bonding part of edge and shear web since it is not accessible to check debonding condition from outside.

Results and discussion As mentioned in Figure 2, we prepared a contour map for blade to be tested before installing to the jig for

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

6

Journal of Intelligent Material Systems and Structures 0(0)

Figure 5. The position of strain gauge attached on the shear web.

static loading test. Totally, three types of input sources, commercial Equo-tip impact tester type-C/D/G, were used to make a contour map database. Figure 6 shows their own contour map for each sensor, which was measured from the input source of D-type Equo-tip tester representatively. The matrix of the database consisted of the energy values which are propagated from each source input to receiving sensor. In other words, these contour maps give unique information for blade such as a fingerprint of the tested blade. Therefore, we can obtain a database for the attenuation characteristics of elastic waves on the blade with complex materials. The red line indicates high value of signal energy, and then level down to blue and black lines according to their values. You can also see the red line contour around each sensor position. Typically, there is higher attenuation in the y-axis direction than in the x-axis direction because there is a material changing and a shear web interrupt in the y-axis direction as shown in Figure 1. In addition, the result shows less attenuation around root part which is made of only GFRP, that is, signal attenuation depends on the material changing in the blade as shown in Figure 6. The thickness of blade has also affected the pattern of contour map. We have found that the signal attenuation was also affected by resonant frequency of used sensor. In case of 30- and 60-kHz resonant frequency sensor, highfrequency sensor showed small difference in each contour line gap. This means that the high-frequency signal attenuates more than low-frequency one. Consequently, if we can obtain some information for the contour line pattern, it will be helpful to understand the material composition of blade by analyzing the contour map. Figure 7 shows a tip deflection and cumulative AE events corresponding to the loading time. It was found that total AE event increased suddenly at around 2600 and 5000 s. This corresponds to the initiation of 60% and 80% of the loading scheme, where it might be a

sudden initiation of damages or failure in the edge of blade. Especially, around 80% of maximum loading, there were a series of big audible sound due to severe damage propagation. We have done a displacement control static test during AE signal acquisition. That is, holding time during static loading test was maintained by fixing the displacement of blade. Therefore, there are some load variations and AE activities during the period of holding displacement. That is, it was shown that AE hit variation occurred during constant tip deflection. This phenomenon actually occurred during static test since sudden damage occurrence in the blade gives rise to the changes in the load and displacement. Figure 8 shows the peak amplitude of each AE event versus loading time. These event data correspond to AE sensor 4 as shown in Figure 4 and were located near the leading edge of the blade. As shown in Figures 8 and 10, the symbol ‘‘x’’ indicates the peak amplitude of located events for damaged area. On the contrary, the symbol ‘‘d’’ represents the results for scattered area. Therefore, it was found that most of AE signals with high amplitude were generated during about 4000–6000 s (70%–90% loading range). However, this amplitude versus time graph shows only the trend of signal intensity for the loading time history. So, we have tried to plot the distribution of signal amplitude for event duration, which is known as a cross-plot for each AE parameter such as amplitude, duration, rise-time, energy, and so on. As shown in Figure 9, you can see the individual feature of AE signals that give us certain information for the shape of the acquired AE signal. In other words, this cross-plot for each different load range shows the correlation of the distribution between AE amplitude and signal duration. Especially, you can see the high-energy AE signal with long duration and relatively high amplitude in Figure 9(c). This result means that the high-energy signals were generated more in the load range of 70%–90% due to damage

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

Han et al.

7

1000 900 800 700 600 500 400 300 200 100 0

450 400 350 300 250 200 150 100 50 0

Tip deflection (mm) Total Hits

0

2000

4000

6000

8000

Total Hits

Tip deflection (mm)

Figure 6. Contour map of 100-kW blade for each sensor (Equo-tip tester D type): (a) sensor 1, (b) sensor 2, (c) sensor 3, (d) sensor 4, (e) sensor 5, and (f) sensor 6.

10000

Time (sec)

Figure 7. Load, blade tip deflection, and accumulated AE event graph of static test. AE: acoustic emission.

Figure 8. AE amplitude versus time for sensor 4 (each event was located near the leading edge). AE: acoustic emission.

such as crack propagation. This phenomenon was confirmed by visual observation during static load testing. On the contrary, low-energy signals with lower amplitude and short duration were always distributed during the whole range from loading to unloading scheme as

shown in Figure 9(a) to (d). These low-energy signals are considered as the AE events generated by the small cracks or debondings. The results of source location using the newly developed energy contour map algorithm are shown in Figure 10. Figure 10 shows both the location of damaged area and the number of located AE events. The red circle indicates highest number of event group, 51–70, which is located in the same area. The other four colored circles show each different number of event group lower than red one. There is a remarkable concentrated location of damages, which is located at around 1100 mm distance from the root. This damaged area is located in the leading edge of the blade. The second highest number of event group is located in the trailing edge of the blade, which is shown by a pink circle in the upper side of the figure. Both of these two damaged areas are located in the part of geometrical change from round root to flat skin, at 1000–1500 mm distance from the root shown in Figure 10. That is, this area has a severe stress concentration in the whole area of the blade. On the contrary, there are scattered source locations that are located at around 2000–4000 mm distance. These random event locations were represented by sky blue circles, which show low signal intensity and few number of location events. These are considered as a small amount of damage, which was not observable visibly. Totally, 437 AE events were obtained during this static test; most of them were located in the actual damaged area. In addition, it was found that there are two visible cracks in the section of leading and trailing edges.

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

8

Journal of Intelligent Material Systems and Structures 0(0)

80 Amplitude (dB)

Amplitude (dB)

80 60 40 20

40 20

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

Duration (msec)

Duration (msec)

(a)

(b)

0.8

1

0.8

1

80 Amplitude (dB)

80 Amplitude (dB)

60

60 40 20

60 40 20

0

0.2

0.4

0.6

0.8

1

0

0.2

0.4

0.6

Duration (msec)

Duration (msec)

(c)

(d)

Figure 9. Cross-plot of peak amplitude versus event duration for sensor 4: (a) 0%–40% loading, (b) 40%–70% loading, (c) 70%–90% loading, and (d) 90% loading–0% unloading.

Figure 10. Results of source location by energy contour map algorithm. AE: acoustic emission.

Figure 11 shows the photographs of the damaged part of the blade after static load testing. It was found that a visibly clear crack is located in the leading edge around 1000–1500 mm far from the root as explained in Figure 10. That is, the result of source location analyzed by the new contour map algorithm shows good agreement with the location of real damage as shown in Figures 10 and 11. Another evidence for the results of source location and damaged area is shown in Figure 12. This figure shows the results of output from strain gage installed on the shear web inside the blade as shown in Figure 5. It was shown that there was an obvious change of strain, especially in the strain gage 2 in the upper side

and in the strain gage 11 in the lower side of the blade as shown in Figure 12. These two strain gages are located on the same position at 2000 mm from the root (1500 mm from the end of shear web). From these results, it was found that most of load was concentrated on the area of 2000 mm distance from the root. Consequently, it was confirmed again that the cracks or damages initiated and propagated from around this high-strain area.

Conclusion This study describes the new concept for identification of damage sources in heterogeneous composite

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

Han et al.

9

Figure 11. Features of damaged area in the blade.

Figure 12. Strain gauge output during static loading test: (a) upper side of shear web and (b) lower side of shear web.

materials and discusses how they can be verified to actual full-scale wind turbine blade during static loading test. In this study, AET was applied to assess the damage in the wind turbine blade during the step-bystep static loading test. In addition, we tried to apply a new source location method which has a new algorithm with energy contour mapping concept. We also measured the deflection of blade tip and the strain of inner shear web in order to analyze the correlation between stress condition and damage identification. The results showed that the AE activities were in good agreement with the stress distribution and damage location in the blade, especially in leading and trailing edges around 1000–1500 mm far from

the root. Finally, the applicability of new source location method was confirmed by comparison of the result of source location and damaged area experimentally. Declaration of conflicting interests The authors declare that there is no conflict of interest.

Funding This work was supported by the Korea Institute of Energy Technology Evaluation and Planning (KETEP) grant funded by the Korean Government Ministry of Trade, Industry, and Energy.

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016

10

Journal of Intelligent Material Systems and Structures 0(0)

References Beattie AG (1997) Acoustic emission monitoring of a wind turbine blade during a fatigue test. In: ASME wind energy symposium/AIAA aerospace sciences meeting, Reno, 6–9 January 1997, NV, p. 239. Borum KK, McGugan M and Brøndsted P (2006) Condition monitoring of wind turbine blades. In: Proceeding of the 27th Risø international symposium on materials science, Rockilde, Denmark, 4–7 Septempber 2006, pp. 139–145. Caprino G, Lopresto V, Leone C, et al. (2011) Acoustic emission source location in unidirectional carbon-fiberreinforced plastic plates with virtually trained artificial neural networks. Journal of Applied Polymer Science 122(6): 3506–3513. Ciang CC, Lee J and Bang H (2008) Structural health monitoring for a wind turbine system: a review of damage detection methods. Measurement Science & Technology 19: 122001. Flemming ML and Troels S (2003) New lightning qualification test procedure for large wind turbine blades. In: Proceedings of the international conference on lightning and static electricity (ICOLSE), Royal Aeronautical Society Blackpool, 16–19 September 2003, pp. 36.1–36.10. Ghoshal A, Sundaresan MJ, Schulz MJ, et al. (2000) Structural health monitoring techniques for wind turbine blades. Journal of Wind Engineering and Industrial Aerodynamics 85: 309–324. Godin N, Huguet S, Gaertner R, et al. (2004) Clustering of acoustic emission signals collected during tensile tests on unidirectional glass/polyester composite using supervised and unsupervised classifiers. NDT & E International 37: 253–264. Grosse CU, Finck F, Kurz JH, et al. (2004) Improvements of AE technique using wavelet algorithms, coherence functions and automatic data analysis. Construction and Building Materials 18(3): 203–213. Hamstad MA, O’Gallagher A and Gary J (2002) A wavelet transform applied to acoustic emission signals: part 1.

Source identification. Journal of Acoustic Emission 20: 39–61. Jørgensen ER, Borum K, Mcgugan M, et al. (2004) Full scale testing of wind turbine blade to failure-flapwise loading. Risø-R-1392(EN) Report, June. Roskilde: Risø National Laboratory. Kirikera GR, Shindea V, Schulz MJ, et al. (2007) Damage localization in composite and metallic structures using a structural neural system and simulated acoustic emissions. Mechanical Systems and Signal Processing 21(1): 280–297. Kirikera GR, Shinde V, Schulz MJ, et al. (2008) A structural neural system for real-time health monitoring of composite materials. Structural Health Monitoring 7(1): 65–83. Oliveiraa R and De Marquesb AT (2008) Health monitoring of FRP using acoustic emission and artificial neural networks. Computers & Structures 86(3–5): 367–373. Schubert F (2004) Basic principles of acoustic emission tomography. In: 26th European conference on acoustic emission testing (EWGAE), Berlin, 15–17 September. Schulz MJ and Sundaresan MJ (2006) Smart sensor system for structural condition monitoring of wind turbines. Subcontract Report. NREL/SR-500-40089, August. Golden, CO: National Renewable Energy Laboratory. Sørensen BF, Jørgensen E, Debel CP, et al. (2004) Improved design of large wind turbine blade of fibre composites based on studies of scale effects (phase 1): Summary report. Risø-R-1390(EN), September. Roskilde: Risø National Laboratory. Yoon DJ and Han BH (2012) Effective AE source location of damages in the wind turbine blade. Review of progress in Quantitative Nondestructive Evaluation 31: 1599–1605. Yoon DJ, Kim YH and Kwon OY (1990) New algorithm for acoustic emission source location in cylindrical structures. Journal of Acoustic Emission 9(3): 237–242. Yoon DJ, Lee SI, Kwon JH, et al. (2005) Characteristics of patch type smart-piezo-sensor for smart structures. Key Engineering Materials 297–300: 2010–2015.

Downloaded from jim.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016