Prediction of Flow Induced Vibration in Shell-and-Tube Heat Exchangers

Marilize van der Hoven Promoters : Prof. P.S. Heyns Prof. K.J. Craig

Layout of Presentation
Introduction Background HTRI analyses CFD analyses Experimental results Comparison of results Conclusion and Recommendations

Introduction
Flow induced vibration can course premature failure in Shell and Tube heat exchangers There are more than 2000 shell-and-tube heat exchangers at SSF If Sasol wants to increase the production they need to increase the flow velocities Large margins of uncertainty associated with existing software (HTFS/HTRI)

Background Natural frequency of straight tube

Assuming that the tubes are fixed at the tubesheets and simply supported at the baffles with equal span length between the baffles. Axial stresses caused by manufacturing procedures or operating conditions like thermal expansion can affect the natural frequency of the tube
Cn EI fn = 4 2 M L e
0. 5

fns

2 P L = f n 1 + a 2 EI

0.5

Excitation mechanisms for tube vibration

Fluid elastic instability Vortex shedding Acoustic resonance Turbulence buffeting

Fluid Elastic Instability

Flow across tubes produces a combination of drag and lift forces
Tube amplitude

1 1 ' Fxj = U 2 DCDj + U 2 DCDj sin( Dj + Dj ) + g 'j 2 2 1 1 ' Fyj = U 2 DC Lj + U 2 DC Lj sin( Lj + Lj ) + h 'j 2 2

Amplitude limited by neighbouring tube

If the fluid velocity (U) is

above the critical value, vibration will occur

Critical velocity Flow velocity

Vortex Shedding
Vortex shedding is the principal excitation mechanism for flow-induced vibration in cross flow, producing alternating forces, which occur more frequently if the flow velocity is increased

Sru f vs = Do

If the vortex shedding frequency and one of the tubes

natural frequencies differ by less than 20% lock in may occur

Acoustic Resonance
Acoustic vibration only occurs when the shell-side fluid is a vapor or a gas Two types of frequencies can be associated with acoustic vibration: Acoustic frequency of the heat exchanger and the acoustic wake shedding frequency of the tube bundles
fa 0.8 < < 1.2 fs

Acoustic resonance

and

0.8 <

fn < 1.2 fs

Tube vibration

Turbulence Buffeting
Turbulent flow contains a wide spectrum of frequencies distributed around a central dominant frequency. This frequency increases as the cross flow velocity increases.
2 Do u c Do 3.05 1 + 0.28 f tb = pl pt pt

The forces associated with turbulence buffeting is motion dependent (Fluid Structure Interaction problem)

Selection of heat exchangers

To select a suitable heat exchanger following criteria were used. The shell- and tube-side flow must be single-phase flow. The shell-side fluid must be a gas with temperature not exceeding 100C. The heat exchanger should not be covered with thermal insulation. It must be possible to vary the load through the heat exchanger. The flow rates, operating pressure and temperature at the inlet and outlet, should be known. The heat exchanger should also have a history of vibration-related problems.

ES208 Tailgas Shell-and-Tube Heat Exchanger

500 1010 T1 4630 S1

CEN 1020 - 6100 Triple segmental baffles (13)

T2 1630

S2 3720

Shell-side 1 Shell pass I.D 1020mm Tailgas (density 21kg/m) Inlet temperature 55C Outlet temperature 47C

Tube-side 6 Tube pass (1100 tubes) O.D 19mm / pitch 26mm Square tube configuration Water Inlet temperature 30C Outlet temperature 43C

HTRI analysis
HTRI analysis calculated:
f n = 28.58 Hz f a = 174.75 Hz

Critical velocity of 2.89 m/s

Mass flow rate associated with: Vortex shedding (fvs) Turbulence buffeting (ftb) Acoustic vortex shedding (fvsa) Acoustic turbulence buffeting (ftba) Inlet (kg/s) 7.950 9.982 46.713 66.134 Middle (kg/s) 8.004 10.054 47.073 66.601

Addition Natural frequencies

HTRI only calculated the natural frequency for the longest unsupported tube length FEA was used to determine the natural frequencies for the 4 unsupported tube lengths
Associated mass flow rate (kg/s) Natural frequency Inlet Middle Turbulence buffeting Inlet frequency Middle Vortex shedding frequency 27.76Hz 7.75 7.80 9.78 9.85 41.57Hz 11.61 11.75 14.58 14.68 59.80Hz 16.70 16.84 20.98 21.12 92.21Hz 25.69 25.87 32.2 32.42

Margin of uncertainty
The following factors that influence the margin of uncertainty in the prediction of flow induced vibration: The temperature and pressure variation at the shell-side inlet of the heat exchanger The tube support assumptions that are used The clearance between the baffle hole and tube as well as tube and baffle corrosion The correlation values used in calculating the natural frequencies of the tubes The flow patterns through the heat exchanger.

CFD Analysis
Simulate middle section of the heat exchanger with periodic boundaries. Create an equivalent porous model Simulate inlet section with porous model to obtain the necessary back pressure and flow patterns at the outlet boundary

CFD Analysis
Middle section Gambit was used to generate and mesh the model. Mesh was adopted and solved in Fluent using a two-equation turbulent energy and dissipation model

CFD Analysis

CFD Analysis

CFD Analysis

Magnitude of cross-flow velocity (m/s)

CFD Analysis
0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

x-position
v bottom = 14656 .1 & 8 & & & & m m m m m x 33836 .3 x 7 + 30566 .5 x 6 13615 .9 x5 + 3105 .82 x 4 4 4 4 4 4 & & & m m m 354.814 x3 + 22.3272 x 2 0.654188 x1 + 0.0132422 4 4 4

Experimental Results
To test whether the predicted vibration actually occurred, vibration measurements were done on the Tail gas heat exchanger at different flow rates.

Vibration Measurements
A record length of 1024 samples was used and twenty averages were taken for every set of measurements. The sampling frequency was 256 Hz and PSDs were calculated for each measurement set.

Strain gauge measurements

Measurements were taken with KFW-5-120-C116L5M2R strain gauges. Noise associated electronic equipment is in the order of 20 micro strain making accurate measurement with only one strain gauge per position impossible. Four strain gauges were used to amplify the strain measured.

Strain gauge measurements

The 28 Hz line coincides with one of the predicted natural frequencies. A slight increase in amplitude is visible at a mass flow rate of 5.5 kg/s.

x 10 6

-7

Amplitude

2 12 10 15 20 8 25 6 30 35 4 40 45 2 50 0

0 10

Frequency (Hz)

Sets

Waterfall plot

Contour plot

Support measurements
Contour plot
40 30 10 20

Mass flow rate (kg/s)

10 0 -10

4 -20 2 -30 -40 0 24 26 28 30 32 34

Frequency (Hz)

Inlet measurements
Measurements were taken with 500mV/g accelerometers. Vibration measured at frequency of 36 Hz

Comparison on Results
Comparison of the HTRI and CFD flow velocities Comparison of natural frequencies for the experimental results and HTRI analyses

Magnitude of cross-flow velocity (m/s)

Vortex shedding average crossflow velocity

14
12
10

75 kg/s

CFD analyses calculated that mass flow rates between 8 kg/s and 75 kg/s in the inlet section, can cause flow-induced vibration due to vortex shedding. This is a very wide range compared to the HTRI predicted range between 6.95 kg/s and 9.72 kg/s In the middle section of the heat exchanger, flow-induced vibration due to vortex shedding can occur from a mass flow rate of 32 kg/s and upwards.
Frequency (Hz)

50 kg/s
6

32 kg/s
4

16 kg/s
2

8 kg/s
0
0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

y-position

HTRI fvs range

CFD fvs range

Mass flow rate (kg/s)

Turbulence buffeting
At the inlet section, the CFD analyses predicted a mass flow range between 9.3 kg/s and 77 kg/s where vibration may occur. The HTRI analyses predicted a flow range between 8.7 kg/s and 11.3 kg/s (using a 20 percent margin of uncertainty)
HTRI ftb range Frequency (Hz)

CFD ftb range

Mass flow rate (kg/s)

At the middle section the CFD analyses predicted that vibration, due to turbulence buffeting, would occur from a mass flow rate of 37 kg/s and upwards

Fluid-elastic instability
The HTRI analyses predicts vibration due to fluid-elastic instability from a mass flow rate of 11.4 kg/s and upwards. The CFD analyses predicted a mass flow range from 14 kg/s and upwards.
tvs range Frequency (Hz)

ttb range

cr (FEI)

Mass flow rate (kg/s)

IF the CFD predicted mass flow ranges calculated for fluid-elastic instability, turbulence buffeting and vortex shedding are combined, flow induced vibration will occur from a mass flow rate of 8 kg/s and upwards. The HTRI analyses predicted what vibration will occur in the heat exchanger from a mass flow rate of 6.95 kg/s and upwards

If the vortex shedding, turbulence buffeting and fluid elastic-instability regions do not overlap, safe operating zones are predicted by the HTRI analyses. The CFD analyses however do not predict safe operating zones after a certain minimum mass flow rate

tvs range ttb range

Frequency (Hz)

Safe operating zones

cr (FEI)

Mass flow rate (kg/s)

Comparison of experimental results

The measured results confirmed that vibration did occur between 27.5 Hz and 28.5 Hz over the measured mass flow range of 3 kg/s to 100 kg/s. The measurements also confirmed that vibration only occurred in certain sections of the heat exchanger. The measurements at the inlet and outlet section of the heat exchanger did not detect vibration at about 28Hz. The measured and HTRI predicted natural frequency differs by less than 4 percent.

Comparison of experimental results (2)

The amplitude of the vibration measurements on the shell and supports are very small. It was not possible to measure the actual tube vibration amplitude, but the HTRI analyses also predicted very small tube vibration amplitudes (0.3 mm). Only vibration at the lowest natural frequency of the tubes was measured. No vibration at the other three natural frequencies was detected.

Conclusion/Future work
By determining the correlation between the inlet velocity and the velocity distribution through the heat exchanger for different tube to pitch ratios, tube configurations, heat exchanger configurations and baffle configurations, more accurate vibration predictions can be made without the use of numerous expensive CFD analyses. In the CFD analyses, the challenging task of simulating the flow in a structure with multiple tubes was tackled through the use of porous sections (calibrated through detail analysis) together with detail resolution of parts of the flow domain. To determine if the forces on the tubes are sufficiently large to cause premature failure of the heat exchanger, fluid-structure interaction analyses and fatigue analyses should be performed. The strain gauge measurements provided good results at lower frequencies, with the added advantage that strain gauges are inexpensive in comparison with accelerometers. Strain gauge measurements can therefore be used to monitor the vibration levels of a heat exchanger.

Conclusion/Future work
To quantify the margin of uncertainty in the prediction of Flow-induced vibration, better correlation values for different tube configurations and heat exchanger configurations are needed. For large and more expensive heat exchangers or where modifications to heat exchangers are required, a CFD analyses provides valuable information to verify vibration problems.