Optics and Lasers in Engineering 49 (2011) 988997

Contents lists available at ScienceDirect

Optics and Lasers in Engineering

journal homepage: www.elsevier.com/locate/optlaseng

Optical measurement of the dynamic strain eld of a fan blade using a 3D scanning vibrometer
C. Vuye a,b,, S. Vanlanduit b, F. Presezniak b, G. Steenackers b, P. Guillaume b
a b

Department of Applied Science and Technology, Artesis Hogeschool Antwerpen, Paardenmarkt 92, B-2000 Antwerp, Belgium Acoustics and Vibration Research Group, Department of Mechanical Engineering, Vrije Universiteit Brussel, Pleinlaan 2, B-1050 Brussels, Belgium

a r t i c l e i n f o
Article history: Received 2 September 2010 Received in revised form 23 January 2011 Accepted 24 January 2011 Available online 21 February 2011 Keywords: 3D-scanning laser doppler vibrometer Strain measurement Strain gages Fan blade

a b s t r a c t
Understanding the origin of the stress and strain distribution is crucial to increase the durability of components under dynamic loading. Numerical simulations based on nite element (FE) models help with this understanding but must be validated by real measured data. Updating the FE model using the measured data is often the next step in the design process. In this paper the recently developed 3Dscanning laser doppler vibrometer (3D-SLDV) is used to measure the 3D-displacement of a fan blade, which is then used to calculate the dynamic strain distributions. The measurement principle and experimental setup are discussed thoroughly. The experimental results are validated by using a FE model on the one hand and strain gage measurements on the other. It is shown that this technique is capable of measuring normal strain far below 1 microstrain. This technique has the potential to ll in the gap of accurately measuring small (full-eld) normal and shear strains at both low and high frequencies, where other optical techniques (and strain gages) would certainly fail. & 2011 Elsevier Ltd. All rights reserved.

1. Introduction A primary goal in many studies of structural dynamics is the identication of the locations on a structure at which failures are most likely to occur. Since it is well known that damage due to vibration is associated with uctuating strain, it is very desirable to identify the locations of maximum strain. To assure dynamic and fatigue strengths it is necessary to measure the dynamic strain distribution. Furthermore, obtaining an accurate and fulleld strain distribution is preferred when using experimental data to verify and update a nite element (FE) model. One of most fundamental methods to measure strain is to use a strain gage. Although very small strains can be measured, there are a lot of drawbacks to this method. The measurement of the strain is limited to the location of the strain gage, which cannot be moved thereafter from point to point. The placement of strain gages is time consuming and in small, lightweight structures strain gages and their connecting cables create an additional mass-loading effect and added damping. Determining the exact position of the strain gage on the test surface can be another difculty, since the strain maxima in real components can deviate from the model-predicted
Corresponding author at: Department of Applied Science and Technology, Artesis Hogeschool Antwerpen, Paardenmarkt 92, B-2000 Antwerp, Belgium. Tel.: + 32 495440707; fax: + 32 32318670. E-mail addresses: cedric.vuye@artesis.be (C. Vuye), steve.vanlanduit@vub.ac.be (S. Vanlanduit).

locations due to manufacturing variations and tolerances. Even with the presence of the FE model that would enable the determination of the theoretical maxima, there is still likely to be some variance in the actual location of the strain maxima. Placing a large number of strain gages to compensate for this is no practical solution either. Some examples where strain gages would denitely fail are high temperature test setups ( 4 500 3 C) or during dynamic high cycle fatigue (HCF) tests [1]. In [2] other difculties when measuring with strain gages such as temperature drift, achieving accuracy below 10 microstrain, the correct attachment to the measurement surface and the problem of measuring at higher modes are discussed further. In the search for a better (strain) measurement technique many optical methods were developed, such as electronic speckle and grating interferometry, pattern interferometry (ESPI), Moire thermoelasticity and photoelasticity, X-ray tomography and digital image correlation (DIC). All these techniques are based on the theory that the strain distribution can be calculated from measuring the displacement eld (see Eq. (1) in Section 2). It is possible to obtain the strain distribution without any optical method using only accelerometers and performing a modal analysis, as shown in [3], but due to the limited number of measurement points only a low spatial resolution can be achieved. The most widely used non-contact optical method for measuring the strain distribution is no doubt digital image correlation (DIC). With DIC, images of a (vibrating) object are acquired in

0143-8166/$ - see front matter & 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.optlaseng.2011.01.021

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

989

digital form and are analyzed to extract sensor-plane motions that can be converted into full-eld measurements on the corresponding object. Both two-dimensional and three-dimensional DIC (2D-DIC and 3D-DIC) [4] are capable of indirectly measuring and visualizing the strain distributions, although these methods are limited to higher strain values. Depending on the desired spatial resolution strains can be measured down to the order of 100 microstrain (low spatial resolution) or down to a few thousand microstrain (high spatial resolution). Furthermore, the 2D-DIC method using a single-camera will give in-plane deformation measurement results as though the object surface was a planar object surface [5]. Therefore, if the test object is of a curved surface, or 3D deformation occurs after loading, the 2D-DIC method is no longer applicable. Another non-contact optical measurement technique determines the strain distribution from vibration measurements using a Scanning Laser Doppler Vibrometer (SLDV). This technique has been in use since the 1990s [1,69], but it should be noted that in these publications only the bending deformation (and strain) could be measured since they were restricted to using single beam Laser Doppler Vibrometry. Moreover, due to poor transducer quality, early applications of this technique required extensive spatial ltering to improve the quality of the (bending) strain estimates, at the expense of the spatial resolution. Strain gages and a SLDV can also be used in a different manner [10]. Here the displacements at low frequencies (up to 100 Hz) were determined using strain gages by rst performing a modal analysis and then using only the measured strain components. The SLDV was used in this scenario only to verify the obtained results. It is not only possible to obtain the strain distribution from SLDV displacement measurements but, in certain cases, it is also possible to achieve the opposite and to obtain the displacements from strain measurements [10]. Recently 3D Scanning Laser Vibrometers, with sufcient spatial resolution and the associated high resolution decoders, have entered the market which allow the non-contact (remote) measurement of not only the out-of-plane displacement component, but also the in-plane displacement components. In this paper the strain distribution of a fan blade is calculated from vibration measurements using a 3D-Scanning Laser Doppler Vibrometer (3D-SLDVa Polytec PSV-400-3D), an accurate strain distribution (including both normal and shear strain) in three dimensions can now be obtained [11,12]. It will be shown in this paper that this technique has the potential to ll the gap of accurately measuring small (full-eld) strains at higher frequencies. The measurements executed with the 3D-SLDV are compared with a FE model on the one hand and strain gage measurements on the other. Measurements were executed at different resonance frequencies (up to 6 kHz) and at different vibration levels to identify the sensitivity of both the strain gages and the 3D-SLDV. This paper is outlined as follows: in Section 2 the basic theory for calculating strain from vibration measurements is explained; in Section 3 the measurement principle and setup are discussed, including the placement and inuence of the strain gages; Section 4 shows the comparison of the measurement results (both 3DSLDV and strain gages) with a FE model. In this section the measured resonance frequencies, mode shapes, normal strain and measurement sensitivity are discussed in detail. General conclusions, merits and limitations of the proposed method will nalize this paper in Section 5.

deformation theory, the strain components in a plate due to bending are given by [9,13]

exx x, y, t z

@2 wx, y, t @x2 @2 wx, y, t @x@y @2 wx, y, t @y2

exy x, y, t z

eyy x, y, t z

with z the transverse distance relative to the center of the plate, w(x,y,t) the out-of-plane displacement, x and y the coordinates of a point along the surface of the plate and t the time. For z plate thickness=2, Eq. (1) yields the strains at the surface of a at surface due to bending. Instead of using the second partial derivative of the out-of-plane displacement a more general approach can be used to determine the strain at the surface of a structure by taking the 1st derivative of the in-plane surface displacement [8]. Surface strains are often of major concern because they typically are greater than internal strains and thus are more likely to lead to failures. The in-plane displacements u and v can only be measured using the 3D-SLDV. The nite difference approach for estimating the strain from the displacement eld obtained from the lasers is similar to the one used in Finite Element Modeling. Other nite difference formulations and possible error sources in nite difference predictions are discussed in detail in [13]. The strain calculation in the PSV StrainProcessor software is based on three-noded triangular elements. The strain nite element shape functions for a threenoded element will now be derived for a planar structure as shown in [11]. This analysis could easily be extended to 3Dstructures. Consider the three-noded triangular element as shown in Fig. 1 with interpolation function shown in the following equation:

fx, y a1 a2 x a3 y

This function represents how a value, such as the displacements u and v, can vary across the element. The real weights ai are a function of the values of f at each of the three nodes (f1 , f2 , f3 ) and are given by [11]

a1 a2 a3

1 a1 f1 a2 f2 a3 f3 2A 1 b1 f1 b2 f2 b3 f3 2A 1 c1 f1 c2 f2 c3 f3 2A 3

with A, the area of the triangular element, given by Eq. (4) and ai, bi and ci given by Eq. (5).

2. Theoretical background Using one SLDV [1,69] only the out-of-plane displacement w of a vibrating plate can be measured correctly. According to small

Fig. 1. Two-dimensional three-noded linear element (source [11]).

990

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

 1   A 1=2 1  1

x1 x2 x3

 y1   y2   1=2x1 y2 x2 y3 x3 y1 x1 y3 x2 y1 x3 y2  y3  4

a1 x2 y3 x3 y2 , a2 x3 y1 x1 y3 , a3 x1 y2 x2 y1 ,

b1 y2 y3 , b2 y3 y1 , b3 y1 y2 ,

c1 x3 x2 c2 x1 x3 c3 x2 x1 5

Eq. (2) can be rewritten more compactly in matrix form

fx, y Nx, yF
where N is the FE shape function vector given by N N1 x, y N2 x, y N3 x, y with Ni 1=2Aai bi x ci y. The interpolation function vector F is given by

6
Fig. 2. Principle of strain calculation (source [14]).

F f1 f2 f3 t
t

with taking the matrix transpose. Now consider the displacement u, vt at an arbitrary location (x,y). This is given by the product of shape function equations at the location and the vector containing the displacements at the nodes: " # ux, y Nx, yU 9 vx, y where the shape function matrix N(x,y) is (dropping the spatial dependence (x,y)) given by " # N1 0 N2 0 N3 0 10 N 0 N1 0 N2 0 N3 The column vector of nodal displacements U is given by 2 3 u1 6 7 6 v1 7 6 7 6 u2 7 6 7 U6 7 11 6 v2 7 6 7 6 u3 7 4 5 v3 The 2D strains are obtained from differentiating the displacement vector: @u @x @v ey @y

After the displacement eld is measured with the 3D-SLDV the strains are calculated using the PSV StrainProcessor software package, which is based on the previously stated equations using three-noded triangular elements. Another simpler way of explaining the measurement principle is by stating that the strain is calculated by placing simulated strain gages at each edge of each triangle as shown in Fig. 2. Strain for each triangle edge can be calculated from the difference in displacement along the edge divided by the length of the edge. For the actual strain calculation, the displacement data are transformed into two in-plane components and one outof-plane component for each surface triangle. Thus, three surface strain values are calculated, two linear strain values ex , ey and one shear strain value gxy . The surface strain values are then transformed back into the global coordinate system.

3. Experimental formulation In this section the measurement principle and some details about the measurement setup are discussed. 3.1. 3D-SLDV principle The 3D-SLDV makes a series of non-contact vibratory velocity measurements on a predened grid, which consists of points and triangular elements, using the laser probes to characterize dynamic strain at each point instead of attaching individual strain gages. The measurement grid can either be imported from a FE model and then be aligned with the real specimen, or it can be dened manually with the geometry of the measurement points. At every point on the measurement grid, the three laser beams are aligned using the socalled VideoTriangulation. This allows a high precision measurement of both the geometry and also the 3D dynamic motion vectors at the measurement points. Thus, the test setup is fast and very repeatable with no mass loading from an attached transducer [14]. The 3D-SLDV operates on the Doppler principle and uses three laser heads to measure the instantaneous vibratory velocity in the direction of each laser beam. One of the potentially limiting factors when using a SLDV is the scattering of the laser light on a rough surface generating speckle noise. This has been an interesting research topic since the 1990s [15] and the problems still occur [16,17]. In this case speckle noise is of less importance due to the fact that:

ex

gxy

@u @v @y @x

12

The relationship between the strains and the nodal displacement vector U can be written as 2 3

ex

6 ey 7 4 5 BU

13

gxy

with B the strain 2 b1 0 6 B 1=24 0 c1 c1 b1

matrix given by 3 b2 0 b3 0 0 c2 0 c3 7 5 c2 b2 c3 b3

14

 the displacements are very small with measured frequencies

well over 100 Hz,

It should be noted that the strains across linear triangular elements are, therefore, uniform and hence this element is known as the constant strain element.

 the use of a special spray to obtain a higher degree of reected

laser light,

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

991

Fig. 3. Measurement of the displacement eld on a at plate and orthogonal transformation (source [11]).

 signal enhancement [18] (speckle tracking) was used during the

measurements which enables the laser beam to be moved slightly from its position to remeasure a point when the signal-to-noise ratio (SNR) is below a certain threshold level (an average SNR of up to 3040 dB was achieved). The necessary displacements can, therefore, be easily calculated from the measured velocities. The displacement components in three orthogonal directions are obtained via an orthogonal decomposition, see Fig. 3. To measure the kinematic variables (such as strain) it is necessary to undertake harmonic measurements using a reference signal since the phase information between nodes is essential. Transfer functions between a reference (usually the source of vibration) and the three displacements measured by the laser heads provide the necessary phase information. Averaging of the transfer functions reduces the inuence of non-correlated motion such as air-borne and groundborne noise. 3.2. Test conguration The rst step before starting a measurement with the 3D-SLDV is the 3D-alignment to obtain a perfect laser beam alignment and geometry matching. The object under test can now be placed in front of the 3D-SLDV in order to create a measurement grid. In this experiment the grid was created using the Polytec PSV software. Alternatively it could also be imported from a FE model. The coordinates of three points from the FE model, see Section 4.1, were used in the PSV software to be able to compare the results from the measurement and the FE model (same global coordinate system). The next step is performing a precise measurement of the coordinates of all grid points using the geometry scan unit and video triangulation. Both the video camera and the three scanning heads are shown in Fig. 4. The used measurement grid, 33 by 33 points, is shown in Fig. 5. The subdivision in triangular elements is also visible in this gure. As test component a fan blade, part of a fan as shown in Fig. 6(a), was selected. The component is made of aluminum, weighs approx. 60 g, has dimensions of approx. 50 by 50 mm and has a thickness ranging between 1 and 5 mm. The part was particularly interesting because of its 3D-curvatures, small size, low weight, high resonance frequencies and expected small strains. One fan blade was equipped with strain gages, see Section 3.3, and 3D-SLDV measurements were made on two different fan blades, one without (blade A) and one with strain gages (blade B), see Fig. 6(b). To reduce unwanted reections on the curved surface by creating a more diffuse light and to obtain a higher degree of reection of the laser light a special
Fig. 4. 3D-SLDV measurement setup: (a) PSV-400 scanning heads, (b) video camera, (c) fan blades mounted on and (d) B&K shaker.

Fig. 5. Measurement grid of 1089 points.

spray was used on the fan blades (ARDROXs by Chemetall). It is a suspension of an inert white powder in a quick drying solvent, which can be easily removed afterwards and has no effect on most common materials of construction. The fan blades were mounted on a shaker (Bruel & Kjaer Vibration Exciter Type 4808) and a frequency sweep was executed to obtain a frequency spectrum, as shown in Fig. 7.

992

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

Fig. 6. (a) Fan blade mounted in complete fan and (b) measurement setup showing blades A and B. Table 1 Comparison of the resonance frequencies. # FE model, f (Hz) 3D-SLDV Blade A (without gages) f (Hz) 1 2 3 4 5 Fig. 7. Frequency spectrum of the fan blade A. 649.84 2014.8 3081.7 4183 5865.4 650 2100 2725 3975 6138 Rel. diff. (%) 0.02 4.23 11.57 4.97 4.65 Blade B (with gages) f (Hz) 650 2188 2725 3725 5950 Rel. diff. (%) 0.02 8.60 11.57 10.95 1.44

The resonance frequencies, see Table 1, were identied by selecting the peaks in the frequency spectrum, visualizing the corresponding mode shapes and comparing them with the FE model. The strain measurements were then executed at these resonance frequencies using a sine excitation at different vibration levels. The signals were generated using the Polytec on-board signal generator and amplied by a B&K Power Amplier (Type 2706). Using the attenuator button it was very easy to obtain vibration levels of 0 dB (10 V), 20 dB and 40 dB, see Section 4.4. The average measuring time was about 1.5 h for the rst measurement including a geometry scan and 45 min for the other measurements.

Fig. 8. Frequency spectrum of the fan blade B with strain gages.

3.3. Strain gages The strain gages (Vishay Micromeasurements & SR-4s ) were placed exactly in the middle of the fan blade B. This way the exact location of the strain gages could be determined more easily in the FE model and 3D-SLDV measurements. The considered type of strain gage was implemented to only measure the normal strain (in the direction of the strain gage grid), not the shear strain. The strain gages were placed in a quarter bridge conguration. The strains were acquired using the programmable quad bridge amplier (PQAB), a strain gage conditioning module, on the LMS SCADAS III data acquisition system. The measurements were executed and processed using the LMS Test.Lab software. The effect of these strain gages can clearly be seen in Fig. 8, where a frequency sweep executed at the fan blade B is shown. It is clear that some resonance frequencies are shifted and that the amplitude of the associated peaks is reduced due to the added damping. Especially the peak round 2 kHz, compare Figs. 7 and 8, has almost completely disappeared. This shows how important a non-contact (optical) method can be for small size structures. Using only strain gages it would be almost impossible to nd all the resonance frequencies and the correct (maximum) strains.

Fig. 9. Fan blade B with two sets of strain gages and their cabling.

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

993

Fig. 9 shows the two sets of 10 strain gages placed on the fan blade B.

4. Validation experiment In this section the experimental results of both the 3D-SLDV and strain gages are compared with a FE model. 4.1. Resonance frequencies The FE model was created using ANSYS Workbench 12 and is meshed using 67 045 nodes and 39 078 tetrahedral elements, as shown in Fig. 10. After the 3D-SLDV measurements the FE model was updated to obtain similar resonance frequencies. The resonance frequencies obtained by the updated FE model and 3D-SLDV are shown in Table 1. As mentioned earlier measurements were executed on both fan blades A and B (without and with gages). It is clear that the placement of the strain gages has a large inuence on the (higher) resonance frequencies.

Since the mode shapes of the FE model match the mode shapes measured with the 3D-SLDV (see Section 4.2), the differences between the resonance frequencies were not relevant so no more effort was put into further updating (the boundary conditions of) the FE model.

4.2. Displacement Figs. 11 and 12 show the excellent resemblance of both the mode shapes from the FE model and the 3D-SLDV. As examples, the 1st and 4th resonance frequency is shown, but the agreement between measurement and FE model at the other resonance frequencies is comparable.

4.3. Normal strain The rst part of this section shows the strain distributions obtained from the 3D-SLDV measurements compared to the FE results. The second part shows a comparison between 3D-SLDV,

Fig. 10. (a) FE model and (b) meshing.

Fig. 11. Mode shape at 650 Hz: (a) FE model and (b) 3D-SLDV.

Fig. 12. Mode shape at 3975 Hz: (a) FE model and (b) 3D-SLDV.

994

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

FE model and strain gages using slices at the location of the strain gages. 4.3.1. Full-eld strain distribution As Fig. 13 shows there is an excellent agreement between the FEM and the 3D-SLDV normal strain measurements. Only the measurement of the normal strain in the direction of the strain gages (Zdirection) is shown for the 1st, 4th and 5th resonance frequency. The results for the other frequencies and other directions are comparable. Only the measurement of the normal strain perpendicular to the curved surface (X-direction) shows some more deviation. Measurements with the 3D-SLDV were also performed on fan blade B (with the strain gages). Using the ARDROXs spray it was even possible to measure the strains on top of the strain gages as shown in Fig. 13(f). 4.3.2. Slices As shown earlier (see Section 3.3) applying the strain gages on the fan blade surface changes its frequency response. A direct comparison between the measured normal strains is, therefore, impossible since the displacement amplitudes and resonance frequencies between the fan blades A and B differ completely. Figs. 1418 will, therefore, only show relative normal strains

(parallel to the Z-axis), where the strains obtained from the strain gages and FE model have been rescaled to the strains measured with the 3D-SLDV on blade A.

Fig. 14. Slice of the normal strain at 650 Hz.

Fig. 13. Normal strain in the Z-directiontop 650 Hz/middle 3975 Hz/bottom 5950 Hz: (a) FEM, (b) 3D-SLDV, (c) FEM, (d) 3D-SLDV, (e) FEM and (f) 3D-SLDV.

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

995

Fig. 15. Slice of the normal strain at 2100 Hz.

Fig. 18. Slice of the normal strain at 6138 Hz.

Fig. 16. Slice of the normal strain at 2725 Hz.

successful, although the measurements were partially executed on top of the installed strain gages. Fig. 15 shows the results at 2100 Hz. The measurements for the rst 10 strain gages failed at this frequency but the second set of strain gages follows both the FE model and 3D-SLDV results perfectly. Both 3D-SLDV results (blade with and without strain gages) follow the FE prediction closely. The result on blade B seems even better than the other one, but it has to be noted that they were performed at different frequencies (2100 Hz for blade A and 2188 Hz for blade B). At 2725 Hz, Fig. 16, the strain gages follow the 3D-SLDV measurements closely, but there is a larger discrepancy between the measured results and the FE prediction. At this frequency no measurement was performed with the 3D-SLDV on blade B. Fig. 17 shows the results at 3975 Hz. Again all measurements agree well to the FE prediction. The rst 10 strain gages seem to differ slightly, but they match the 3D-SLDV results for blade B perfectly. Again it has to be noted that the measurements with the 3D-SLDV were done at different frequencies (3975 Hz on blade A and 3725 Hz on blade B). Finally Fig. 18 shows the results at 6138 Hz. At this frequency the differences are the largest. A slight change in resonance frequency (e.g. due to the placement of the strain gages or small differences between the two used blades) will cause a large difference in the strain results, which may account for these noticeable differences. For an easier comparison the normalized root mean square error (NRMSE), as dened by Eq. (15), is calculated and shown in Table 2. The measurement with the 3D-SLDV was taken as reference. NRMSE sP Nk 1 SLDV nenj2 n 0 je jeSLDV j2 15

Fig. 17. Slice of the normal strain at 3975 Hz.

Fig. 14 shows the results at 650 Hz. It is clear that the measurements match the simulation quite well and that the strain gages follow the FE prediction perfectly. Surprisingly enough the 3D-SLDV measurement on blade B was also

with e, respectively, the FE prediction or strain gage results and Nk the total number of data points in common between, respectively, 3D-SLDV and FE model or 3D-SLDV and strain gages. Overall it can be concluded that the 3D-SLDV measurement results match both the FE prediction and the strain gage results closely. The measured maximum normal strains are in the range of 5220 me, where other optical techniques would denitely fail. Table 3 shows a comparison of the measured maxima and minima, on the one hand obtained with the strain gages and on the other hand the 3D-SLDV strain results on blade B. This shows

996

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

again that the measured strain results on blade B correspond quite well with the strain gage measurements. 4.4. Sensitivity In order to investigate how sensitive the 3D-SLDV strain measurement setup is, all measurements on blade A were executed at lower vibration levels. Figs. 19 and 20 show the normal strain distribution (Z-axis) on blade A for the 1st and 4th resonance frequency at different vibration levels (reference level 0, 20 and 40 dB). Each attenuation of 20 dB causes a change in the normal strains of a factor 10. A vibration level of 20 dB is not a problem for the 3D-SLDV (with a measured maximum of 2:56 me at 650 Hz and 0:84 me at 3975 Hz). Only at 40 dB the measured strains become so small (with a measured maximum of 0:22 me at 650 Hz
Table 2 Normalized RMS errors between 3D-SLDV, FE model and strain gages. f (Hz) 3D-SLDV vs FE model NRMSE (%) 650 2100 2725 3975 6138 1.12 2.68 5.96 0.60 24.10 3D-SLDV vs strain gages NRMSE (%) 1.87 3.27 7.81 3.42 11.28

and 0:08 me at 3975 Hz) that the results become a little more noisy, especially at the tip of the blade. It can be concluded that the 3D-SLDV can accurately measure normal strains far below 1 me. The results from the strain gage measurements show that the strain gages can only measure above 1 me. Due to the added damping on blade B, however, this means that only the measurements at the reference level showed reliable results for the strain gages (see also Table 3 for the measured maxima at each resonance frequency). Only at 650 Hz the strains were high enough to obtain reliable results at 20 dB.

5. Conclusions As shown in this paper it is possible to obtain reliable dynamic surface strains from 3D-displacement data obtained with a 3DSLDV. Normal strains can be measured accurately as is shown by comparing the 3D-SLDV measurement results with both a FE model and strain gage measurements. The comparison for the shear strain between 3D-SLDV measurement and FE model gave similar (satisfying) results but they were not included in this paper due to the lack of shear strain gage measurements for comparison. It is shown that applying strain gages on the fan blade inuences the dynamic behavior of the structure. Some resonance frequencies are shifted and the resonance peaks are signicantly reduced at certain resonance frequencies. The 3D-SLDV noncontact measurements are not clearly inuenced by these disadvantages. Moreover, it is shown that the sensitivity of the 3DSLDV is much larger than of the strain gages. It is possible to measure (normal) strains up to one order of magnitude smaller compared to strain gages. The only minor shortcoming of this measurement technique seems to be the less accurate results perpendicular to the surface for both the normal strain (Xdirection) and the shear strain (XY-direction). This technique is of course limited to dynamic strain and the cost of the 3D-SLDV can also be called a drawback compared to other optical techniques. The 3D-SLDV certainly has the potential to ll in the gap of accurately measuring small (full-eld) normal and shear strains

Table 3 Comparison of the maximum and minimum measured normal strain me. f (Hz) Strain gages Max 650 2188 3725 5950 24.77 1.06 1.39 0.97 Min 0.47 0.10 0.14 1.86 3D-SLDV blade B Max 28.95 0.43 2.97 0.94 Min 1.68 0.06 0.20 3.38

Fig. 19. Normal strain distribution at 650 Hz at different vibration levels: (a) reference 0 dB, (b) 20 dB and (c) 40 dB.

Fig. 20. Normal strain distribution at 3975 Hz at different vibration levels: (a) reference 0 dB, (b) 20 dB and (c) 40 dB.

C. Vuye et al. / Optics and Lasers in Engineering 49 (2011) 988997

997

at both low and high frequencies, where other optical techniques would denitely fail.

Acknowledgments This research has been sponsored by the Flemish Institute for the Improvement of Technological Research in Industry (IWT) and the Fund for Scientic Research (FWO). The authors also acknowledge the research council of both the Artesis Hogeschool Antwerpen and the Vrije Universiteit Brussel, and Polytec GmbH for the opportunity to use the 3D-SLDV for these measurements. The authors would like to thank Thomas Mechnig and Jean-Paul Schepens for their assistance during the measurements and the meticulous preparation of the strain gages. Finally the authors wish to thank the reviewers for their insightful and constructive comments that helped improve this paper.

References
[1] Camden M, Simmons L. Using a laser vibrometer for monitoring dynamic strain, modal analysis and calculating damping. In: Proceedings of SPIE, Proceedings of the conference on laser interferometry IX, vol. 3479; 1998. [2] Foss G, Haugse E. Using modal test results to develop strain to displacement transformations. In: Proceedings of SPIE, Proceedings of the 13th international modal analysis conference, vol. 2460; 1995. [3] Okubo N, Yamaguchi K. Prediction of dynamic strain distribution under operating condition by use of modal analysis. In: Proceedings of SPIE, Proceedings of the 13th international modal analysis conference, vol. 2460; 1995.

[4] Sutton M, Yan J, Tiwari V, Schreier H, Orteu J. The effect of out-of-plane motion on 2D and 3D digital image correlation measurements. Optics and Lasers in Engineering 2008;46:74657. [5] Pan B, Qian K, Xie H, Asundi A. Two-dimensional digital image correlation for in-plane displacement and strain measurement: a review. Measurement Science & Technology 2009;20(6):062001. [6] Miles R, Bao W, Xu Y. Estimation of random bending strain in a beam from discrete vibration measurements. Journal of Sound and Vibration 1994;174(2): 1919. [7] Moccio C, Miles R. Measurement of the transfer function between bending strain and a transient load using a scanning laser vibrometer. Journal of Sound and Vibration 1996;189(5):6618. [8] Xu Y, Miles R. Experimental determination of bending strain power spectra from vibration measurements. Experimental Mechanics 1996;36(2):16672. [9] Xu Y, Miles R. Full-eld random bending strain measurement of a plate from vibration measurement. Journal of Sound and Vibration 1996;191(5):84758. [10] Pisoni A, Santolini C, Hauf D, Dubowsky S. Displacements in a vibrating body by strain gauge measurements. In: Proceedings of SPIE, Proceedings of the 13th international modal analysis conference, vol. 2460; 1995. [11] Cazzolato B, Wildy S, Codrington J, Kotousov A, Schuessler M. Scanning laser vibrometer for non-contact three-dimensional displacement and strain measurements. In: Proceedings of acoustics 2008; 2008. [12] Retze U, Schuessler M. Dynamic stress and strain measurement. Technical Report, Polytec; 2010. [13] Karczub D, Norton M. Finite differencing methods for the measurement of dynamic bending strain. Journal of Sound and Vibration 1999;226(4): 675700. [14] Polytec. Polytec data sheetoptical measurement of dynamic stress and strain. Technical Report, Polytec; 2009. [15] Strean R, Mitchell L, Barker A. Global noise characteristics of a laser Doppler vibrometerI. Theory. Optics and Lasers in Engineering 1998;30:12739. [16] Rothberg S. Numerical simulation of speckle noise in laser vibrometry. Applied Optics 2006;45(19):452333. [17] Martin P, Rothberg S. Introducing speckle noise maps for laser vibrometry. Optics and Lasers in Engineering 2009;47:43142. [18] Castellini P, Martarelli M, Tomasini E. Laser doppler vibrometry: development of advanced solutions answering to technologys needs. Mechanical Systems and Signal Processing 2006;20:126585.