Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Laboratory Investigation of Long Riser VIV Response

Jaap J. de Wilde and Ren H.M. Huijsmans


Maritime Research Institute Netherlands (MARIN) Wageningen, The Netherlands

ABSTRACT
In order to address the response of a long cylindrical riser in current, design codes (like Vandiver, 1983) require lift and drag coefficients of 2-D sections. Uncoupling the flexible response from the loading these design codes will determine the flexural response based on sectional stiffness and mass coefficients. However the physics of VIV behavior indicate that the hydroelastic response is significant. Therefore a study has been initiated to research the 3-D response of a part of a long riser in current by experiments. A circular steel pipe of 16 mm diameter and 12.6 m long was towed horizontally in a towing tank at speeds between 0.5 and 3.0 m/s and with pretensions between 0.5 and 2.5 kN. Measurements were made of the drag loads, the acceleration and the bending moments. State-of-theart fibre optical measuring techniques were deployed to obtain detailed insight into the complex VIV response of the test pipe. The first analysis of the test data shows a very complex response, including strong coupling between cross flow and in-line motions, beating, multimode response, traveling wave response, etc. Further analysis will continue and additional results will be published later.

another or from one set of modes to another, even in steady uniform current. Non-stationary response may occur as well, including mode swapping or traveling waves. Full scale measurements, model tests and theoretical considerations indicate that all of these types of behavior can occur for deep water risers, especially in sheared current situations. For the riser design and the VIV analysis it is of key importance to understand when single mode response dominates or when several modes will be excited simultaneously. High quality experiments showing high mode VIV behavior of long risers or cables are rare (Hong et al., 2002; Kleiven, 2002 and Larsen, Vandiver, Vikestad, and Lie, 1997), especially at high Reynolds numbers (Allen and Henning, 1997; Allen and Henning, 2001 and Simantiras and Willis, 1999). Prototype measurements on real risers are presently undertaken, but results have not been published yet in open literature. Such measurements as well as model scale experiments are much needed for the better understanding of the high mode response behavior and for the validation of semi-empirical VIV prediction tools, as well as the latest developments in CFD. Laboratory experiments require a large test facility to accommodate sufficient riser length, even for relatively low Reynolds numbers and relatively small L/D ratios. Moreover a strong and stiff carriage is needed to cope with the large drag loads and pretensions. A large amount of instrumentation is required to capture the complex spatial VIV response of a long riser in detail. The pipe motions of every possible mode need to be measured in at least two degrees of freedom (cross flow and in-line) with sufficient accuracy and at a sufficient sampling rate. Obviously, the instrumentation and cabling should be non-intrusive and preferably be placed inside the tests pipe.

KEY WORDS: Vortex; vibration; VIV; riser; high mode; experiments; fibre optics.

INTRODUCTION
One of the great challenges in the offshore industry is still the assessment of the motions of a circular cylinder in waves and current for application to risers or riser bundles in water depths up to 3,000 m (approximately 10,000 feet). Here the fatigue life of the riser is often dominated by Vortex-Induced Vibrations (VIV). Also the important concern of riser interference is largely governed by VIV. There is a great difficulty in predicting the response of a long span of riser in current. The riser may respond in several possible modes, but it is most often uncertain whether the riser will predominantly respond in one or a few modes or that it will respond in several modes simultaneously (Blevins, 2001; Triantafyllou et al., 1999 and Vandiver, 1983 and 1993). The response may also transit from one mode to

TEST DESCRIPTION Model Basin


The experiments were conducted in MARINs Shallow Water towing tank in Wageningen, The Netherlands. The basin is 15.8 m wide by 220 m long and is equipped with an overhead carriage. The water depth in the basin is 1.15 m. The carriage has a velocity range of 3 m/s and a power supply of 4 x 15 kW. The horizontal test pipe was supported by

Paper No. 2004-EF-005

De Wilde

1 of 6

two vertical struts from the carriage. A threaded end was used for adjusting the initial pretension in still water. Some flexibility was built into the vertical struts to reduce the increase of the mean tension in the pipe during towing and also to reduce the dynamic variations in pipe tension. The maximum measured in-line deflection of the test pipe was less than 1.0 m or less than 8% of the pipe length.
Carriage

Instrumentation and Data Acquisition


The carriage speed was measured by means of an encoder on a fifth wheel on the carriage. The drag loads were measured on both pipe ends by means of strain gauge type load cells. The vertical (lift) loads and the axial loads were measured in the same way. The instrumentation was calibrated and checked for linearity prior to the testing. Small two-component accelerometers were mounted inside the pipe at two location, respectively at 6.04 m and 9.32 m from the right end of pipe, excluding the 50 mm of the universal joint. Special steel houses were manufactured for each location, in which two accelerometers (x and z) could be firmly mounted. The pipe was cut at the two locations and the steel houses were welded in between. The steel houses were kept small as possible to minimize the influence on the pipe properties. However, in future correlation studies it might still be necessary to take the locally different mass and bending stiffness properties into account. The electrical cables of the accelerometers were fed through the pipe. All signals were sampled at 250 Hz.

Test pipe

Water level

Rail

Tensioner

Transducer

1.15 m 15.8 m

Fig. 1. Test set-up for towing 12.6 m horizontal pipe (cross section)

Test Pipe and Test Set-Up


The test pipe was a seamless thin walled steel pipe of 16 mm outer diameter and 0.8 mm wall thickness. Steel end caps were welded on each end for mounting the pipe between the two vertical struts. A universal joint was placed on both ends to obtain a pinned connection without bending moments, adding about 2 x 50 mm to the total effective pipe length from joint to joint. The load cells were located between the universal joints and the struts. The pipe was filled with tank water.

Fibre Optic Measuring Technique


The right-hand side of the test pipe was instrumented with 40 optical fibre strain gauges. Four fibres of approximately 0.3 mm diameter were mounted in small axial grooves in the pipe surface, at 90 degrees angle: top, bottom, fore and aft. The small grooves were carefully machined in the pipe surface and filled with resin after the fibres had been placed inside, to keep a smooth pipe surface. A total of 10 Fibre Bragg Gratings (FBG) were burned in each fibre, using a spatially varying pattern of intense UV light. A specialized company in UK was responsible for the preparation of the optical fibres and the mounting in the pipe surface. The FGB strain gauges were located at 0.375, 0.750, 1.125, 1.500, 1.875, 2.250, 2.625, 3.000, 3.750 and 4.500 m from the right end of the test pipe, excluding the 50 mm of half a universal joint.
6.6 m 6m

4.500

3.750

3.000

2.625 2.250 1.875

1.500 1.125 0.750

0.375

Fig. 2. Test pipe with load cells and universal joint The properties of the test pipe are summarized in the next table at model scale. The pipe surface was reasonably smooth. Table 1. Properties of test pipe Parameter Diameter Length Axial stiffness Bending stiffness Mass ratio Symbol D L EA EI m+ Unit mm m N Nm2 Value 16 12.6 8.02E6 232 2.29

FBG locations

16 mm

Cross section

Fig. 3. Instrumentation of test pipe with 40 FGB strain gauges Each Bragg grating serves as a wavelength selective mirror, which reflects light with a wavelength corresponding to the so-called Bragg wavelength lB (Doyle, 2003). The rest of the light propagates down the fibre uninterrupted. Physical or mechanical properties such as temperature or stress may affect the reflected wavelength, which in our case was deployed to construct an optical strain gauge. Because of the very constant temperature of the water in the basin (less than 1 degree Celsius variation), the temperature effect could be disregarded and the measured shift in wavelength was as a direct measure of the local strain.

The effective lateral stiffness of the vertical struts was 51 kN/m and acts as axial springs on both ends of the test pipe. The test pipe itself had a total axial stiffness of 640 kN/m, which is much higher. The effective horizontal and vertical stiffness of the end supports are very high.

Paper No. 2004-EF-005

De Wilde

2 of 6

14

Optical fibres are flush mounted

.4 m m

0.3 mm

cladding

n y n = A sin L

(2)

in which y is the co-ordinate along the cylinder length and A the amplitude of the oscillation. The bending moment is proportional to the curvature and can be found by double differentiation of the shape:

M = EI

2n y
2

A 2 n2 L
2

n y sin L

(3)

Bragg grating burned in core of optical fibre

Fig. 4. Schematic drawing of Bragg grating in optical fibre A W3/4250 FBG unit was used for the interrogation of the 40 optical strain gauges. The sampling rate was 250 Hz. The unit illuminates the four fibres with a swept-wavelength light source and measures the reflected light by means of a photodiode detector. Wavelength-division multiplexing (WDM) is used to address the different sensors in each line.

The bending moment is also proportional the difference in strain on two opposing sides of the circular test pipe. In our case the measured differences in strain "Top - Bottom" and "Fore - Aft" were regarded as a measure of respectively the cross-flow and in-line curvature at each location. Using of a Taylor series expansion the displacement at a position y can be expressed by:

z( y ) = z( y0 ) + ( y y0 )

z 1 2 z + 2 ( y y0 )2 + ... y y 2

(4)

TEST RESULTS AND DISCUSSION Theoretical VIV Response of Tensioned Beam


The natural frequencies of a tensioned beam are given by the following relation (Timoshenko, Young and Weaver, 1974):

This relation shows how the displacements can be derived from the measured curvatures. This step, involving an inverse matrix operation for each time step, has not been carried out yet.

Drag Loads
In Fig. 5 the drag coefficient of the vibrating cylinder is plotted as a function of the tow speed. The mean tension in the experiments increased with the tow speed from 0.45 kN in still water to 2.5 kN at the maximum tow speed of 3.0 m/s. The Reynolds numbers are all in the sub-critical regime (7,000 < Re < 43,000).
Drag load

fn =

1 2

n4 + n 2

L2T

EI

EI

( m + ma ) L4

(1)

The mode number (n) represents the number of half waves in the response shape. The other parameters are the pipe length (L), the axial tension (T), the bending stiffness (EI), the mass per unit length (m) and the added mass (ma). Our test pipe of 16 mm diameter had a mass in air of 0.461 kg and an added mass of 0.201 kg per unit length (assuming Cm = 1 in still water). The natural frequencies of the first 12 modes at 1.0 kN pretension are given in Table 2. The vortex shedding frequency at 1.0 m/s current speed is 12.5 Hz, assuming a Strouhal number of 0.2. Lock-in VIV response can therefore be expected for mode 6 or 7 at this speed. Table 2. Natural frequencies of test pipe in still water at 1.0 kN pretension Mode 1 2 3 4 5 6 Frequency 1.55 Hz 3.17 Hz 4.92 Hz 6.85 Hz 9.00 Hz 11.41 Hz Mode 7 8 9 10 11 12 Frequency 14.11 Hz 17.12 Hz 20.45 Hz 24.11 Hz 28.12 Hz 32.48 Hz

3.0000 2.5000 2.0000

Cd [-]

1.5000 1.0000 0.5000 0.0000 0.000

0.500

1.000

1.500

2.000

2.500

3.000

3.500

Tow speed [m/s]


Positive speed Negative speed

Fig. 5. Measured drag coefficient as a function of the tow speed The vibrations of the cylinder lead to higher drag loads than for the same cylinder kept stationary at the same speed, which is known as drag amplification. The drag amplification factor in our tests is about 2 at small speeds and decreases for increasing speeds. At 3.0 m/s the drag coefficient of about 1.37, which is only slightly above the normal drag coefficient for a stationary smooth cylinder of Cd = 1.2. It should be noted that no corrections were made for the end effects and the drag on the universal joints.

The mode shape of an oscillating uniform beam can be represented by a sinusoid:

Paper No. 2004-EF-005

De Wilde

3 of 6

Motion Response
Time traces of the measured motions at the pipe mid are presented in Fig. 6. These motions were derived by double integration of the measured accelerations. The tow speed was 1.0 m/s and the mean tension 0.9 kN.
9.8 Hz 9.8 Hz 19.6 Hz

AX1

AZ1

Fig. 6. Time traces of measured motions at pipe mid The amplitude ratios are respectively A/D = 0.35 for the cross flow oscillations and A/D = 0.64 for the in-line oscillations. A trajectory plot of the cross flow versus in-line motions is presented in Fig. 7.

Fig. 8. Spectral density plot of measured accelerations at pipe mid The unexpected response will be further investigated after submission of this paper. The following comments seem appropriate at this stage however: The sign convention of the signals in Figs. 6, 7 and 8 has been carefully checked and no error was found. The cabling of the accelerometers was checked prior to the test, by rotating the pipe in still water and measuring the gravity component. The stiffness of the vertical struts is 51 kN/m at each side, which is much lower than the axial stiffness of the steel pipe itself (640 kN/m), but may still be high enough to induce significant dynamic tension in the pipe for the higher response modes and response amplitudes. The measured dynamic tension in the presented test 101008 was approximately 10% of the mean tension. The mean tension provided the main restoring mechanism. The contribution of the bending stiffness was relatively small. Dynamic end effect, complex non-linear response or strong coupling between cross-flow and in-line motions may also be responsible for the unexpected results. It seems worthwhile to check for possible Mathieu instabilities (Hagedorn, 1988).

+z Flow +x

Fig. 7. Trajectory plot of cross flow vs. in-line motions at pipe mid A "lying banana" type motion is observed. In general "figures of eight" or "standing bananas" type trajectory plots are expected for the vortexinduced vibrations of a circular pipe. Our unexpected result deserves further investigation. Possibly the dynamic tension in our experiment was higher than for most real marine risers. The trajectory plots of the curvature, presented at the end of this paper, show unexpected complex responses at the different locations of the pipe as well. Spectral density plots of the measured accelerations are presented in Fig. 8. Response peaks can be observed at 9.8 and 19.6 Hz, corresponding to Strouhal numbers of respectively 0.16 and 0.31. The higher frequency is twice the lower frequency. It is assumed that these frequencies are related respectively to the cross-flow excitation (two opposing cross flow vortices per cycle) and the in-line excitation (two in-line vortices per cycle). The theoretical calculated natural frequencies of mode 5 and 6 are close to the lower frequency peak in the spectral density plot. It should, however, be noted that the natural frequencies of lock-in VIV can significantly deviate from the theoretical values, due to the large variability of the added mass.

The dynamic tension is not normally considered in the response analysis of marine risers. Calculations are often based on the linearized beam equation:

2 y t 2

+b

y y 2 2 y T + EI 2 = L( x,t ) t x x x 2 x

(5)

Finally it should be noted that the scaling of our experiment to prototype dimensions is not straightforward. Froude scaling is not the obvious choice, because gravity is not an important parameter in the experiment. Instead it is proposed to scale as follows: geometrical scale: velocity scale: tension: l:1 1:1 l2:1 (e.g. 50:1) (e.g. 2500:1)

Paper No. 2004-EF-005

De Wilde

4 of 6

Trajectory Plots of Curvature Derived from FBG Measurements


Trajectory plots of the measured curvature are presented in Fig. 10. The curvature in z direction is plotted against the curvature in x direction. The 10 plots correspond to the 10 FBG strain gauge locations along the pipe, as shown in Fig. 9. Presented is about 2 seconds of the test.

The observed "lying banana" trajectory at the pipe mid is intriguing and deserves further investigation. In general "figures of eight" or "standing bananas" type trajectory plots are found for the vortex-induced vibrations of a circular pipe. Possibly the relatively high dynamic tension plays a role in our experiments. Further analyses will be carried out to derive the motion response from the measured curvature at the 10 locations of the optical strain gauge measurements. Next it will be attempted to identify the VIV vibration modes from the experiments. Finally it seems worthwhile to validate semi-empirical VIV prediction programs against our test results.

REFERENCES
Allen, DW and Henning, DL (1997). Vortex-Induced Vibration Tests of a Flexible Smooth Cylinder at Supercritical Reynolds Numbers, Proceedings of the International Offshore and Polar Engineering Conference, Honolulu, USA. Allen, DW and Henning, DL (2001). Prototype Vortex-Induced Vibration Tests for Production Risers, Offshore Technology Conference, Paper OTC 13144, Houston, USA. Blevins, RD (2001). Flow Induced Vibrations, Second Edition, Krieger Publishing Company, Malabar, Florida. Brida, D and Laneville, A (1993). Vortex-Induced Vibrations of a Long Flexible Cylinder, Journal of Fluid Mechanics, Vol 250, pp 481-508. Doyle, C (2003). An Introduction to Bragg Gratings and Interrogation Techniques, Smart Fibres. Hagedorn, P (1988). Non-linear Oscillations, Clarendon Press, Oxford. Hong, S, Choi, YR, Park, JB, Park, YK and Kim, YH (2002). Experimental Study on Vortex-Induced Vibration of Towed Pipes, Journal of Sound and Vibration, Vol 249, pp 649-661. Kleiven, G. (2002). Identifying VIV Vibration Modes by Use of the Empirical Orthogonal Functions Technique, OMAE2002-28425. Larsen, CM, Vandiver, JK, Vikestad, K and Lie, H (1997). Vortex Induced Vibrations of Long Marine Risers - Experimental Investigations of Multi-Frequency Response, BOSS, pp 455-468. Marcollo, H and Hinwood, J (2002). Mode Competition in a Flexible Cylindrical Riser, International Mechanical Engineering Congress and Exposition. Simantiras, P and Willis, N (1999). Investigation on Vortex Induced Oscillations and Helical Strakes Effectiveness at Very High Incidence Angles, Proceedings of the International Offshore and Polar Engineering Conference, Brest, France. Timoshenko, S, Young, DH and Weaver, W (1974). Vibration Problems in Engineering, Fourth Edition, John Wiley & Sons. Triantafyllou, MS, Triantafyllou, GS, Tein, D and Ambrose, BD (1999). Pragmatic Riser VIV Analysis, Offshore Technology Conference, Paper OTC 10931, Houston, USA. Vandiver, JK (1983). Drag Coefficients of Long Flexible Cylinders, Offshore Technology Conference, Paper OTC 4490, Houston, USA. Vandiver, JK (1993). Dimensionless Parameters Important to the Prediction of Vortex-Induced Vibration of Long Flexible Cylinders in Ocean Currents, Journal of Fluids and Structures, Vol 7, pp 423455. Willden, R and Graham, G (2002). Multi-Modal Vortex-Induced Vibrations of a Vertical Riser Pipe Subject to a Uniform Current Profile, Conference of Bluff Body Wakes and Vortex-Induced Vibrations BBVIV3, pp 229-232.

10

Fig. 9. Locations of curvature trajectory plots The response was very different at each of the 10 locations, which was also expected for the long pipe. Not expected, however, was the complex shape of the trajectories. Similarly as for the motions in Fig. 7, the curvature response does not resemble much of a "figure of eight". However some cyclic repetition can still be recognized in plots. Plotting the same graphs for a shorter duration shows that approximately the same path is followed every cycle, where each cycle consists of a number of distinct lobes. Swapping from one type of motion to another does not occur in the presented part of the test. Because the curvature is presented and not the actual displacements the results cannot be compared directly with "figure of eight" type plots. However, the plots in Fig. 10 still show how complex the VIV response of the long riser can be, even in uniform current. Further analysis is needed to get more insight in this complex behavior. One of the future aims is to learn more about the mode participation and to compare our measurements with the results of VIV prediction tools. The analysis techniques described by Kleiven (2002) for identifying the VIV vibration modes by use of the Empirical Orthogonal Functions technique will be used.

CONCLUSIONS
The vortex-induced vibrations of a horizontal cylindrical pipe in a cross flow situation were investigated at small scale in a tow tank for a range of tow speeds and pretensions. The slender test pipe with a length over diameter ratio of 788 exhibited VIV response with a high modal content. Conventional accelerometers and state-of-the-art optical fibre stain gauges were successfully deployed for capturing the complex response behavior. Based on the preliminary results presented in this paper, it can be concluded that the vibrational response of the long pipe is much more complex than expected. Single mode response with a stable sinusoidal mode shape was not observed, at least not for the tests with the higher tow speeds. For presented test in this paper (1.0 m/s and 0.9 kN) it is assumed that mode 5 or 6 was dominant, but also much higher modes and double frequency modes (mode 11 or 12) participated.

Paper No. 2004-EF-005

De Wilde

5 of 6

Test 1001008
500 400
Strain Top - Bottom [microstrain] Strain Top - Bottom [microstrain]

Test 1001008
500 400
Strain Top - Bottom [microstrain]

Test 1001008
500 400 300 200 100 -500 -300 0 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

300 200 100 0 -100 -100 -200 -300 -400 -500


Strain Fore - Aft [microstrain]

300 200 100 0 -500 -300 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

0.375 100 300 500

-500

-300

0.750 100 300 500

1.125 100 300 500

Test 1001008
500 400
Strain Top - Bottom [microstrain] Strain Top - Bottom [microstrain]

Test 1001008
500 400
Strain Top - Bottom [microstrain]

Test 1001008
500 400 300 200 100 -500 -300 0 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

300 200 100 0 -500 -300 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

300 200 100 0 -500 -300 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

1.500 100 300 500

1.875 100 300 500

2.250 100 300 500

Test 1001008
500 400
Strain Top - Bottom [microstrain] Strain Top - Bottom [microstrain]

Test 1001008
500 400
Strain Top - Bottom [microstrain]

Test 1001008
500 400 300 200 100 -500 -300 0 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

300 200 100 0 -500 -300 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

300 200 100 0 -500 -300 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

2.625 100 300 500

3.000 100 300 500

3.750 100 300 500

Test 1001008
500 400
Strain Top - Bottom [microstrain]

300 200 100 -500 -300 0 -100 -100 -200 -300 -400 -500
Strain Fore - Aft [microstrain]

4.500 100 300 500

Fig. 10. Curvature trajectory plots derived from FBG strain gauges

Paper No. 2004-EF-005

De Wilde

6 of 6

You might also like