Acoustically Induced Vibration Mitigations in Compressor Piping Systems
Acoustically Induced Vibration Mitigations in Compressor Piping Systems
Acoustically Induced Vibration Mitigations in Compressor Piping Systems
GT2016
June 13 – 17, 2016, Seoul, South Korea
GT2016-57800
ABSTRACT
Acoustically induced vibration (AIV) is a high-frequency NOMENCLATURE
vibration phenomenon that can occur downstream of pressure-
reducing devices such as control valves, restriction orifices, and Symbols
pressure relief or safety valves in compressor piping systems.
These vibrations can lead to high cycle fatigue failures of a Internal pipe radius [m]
downstream piping at side branches or welded supports. c Sound speed [m/s]
Existing methods for screening and analyzing acoustically
Di Internal pipe diameter [m]
induced vibration are not well-grounded in the underlying
physics and thus do not provide a methodology for evaluating a FL Liquid pressure recovery factor (from IEC) [-]
variety of mitigation strategies. Modeling of acoustically LP Sound pressure level [dB ref. 20µPa]
induced vibration is computationally challenging, as it requires LW Sound power level [dB ref. 1pW]
the interaction between tens or hundreds of higher-order
acoustic modes with a similar number of piping shell modes. P1 Pressure upstream of CV [MPa]
In order to obtain better insight into the underlying physics P2 Pressure between CV/PSV [MPa]
of AIV and to characterize the effectiveness of several P3 Pressure downstream of PSV [MPa]
mitigation methods, full-scale blowdown testing was performed
at Southwest Research Institute. Tests were performed using W Sound power [W]
20 MPa nitrogen gas vented at 28 kg/s through a 3x4” pressure ρ Density [kg/m3]
safety valve and multiple header pipe sizes ranging from 12” to
36”. Test configurations included baseline piping geometry at
each size and several AIV mitigations including stiffening INTRODUCTION
rings, viscous damping wrap, and internal acoustic mode Acoustically-induced vibration (AIV) refers to
disruptors. Test results from strain gauges, accelerometers, and high-frequency pipe wall vibration that is excited by an
dynamic pressure transducers show a broadband multimodal acoustic source. Typically, the source of this acoustic excitation
response with dynamic stresses up to 3 kHz near the safety is an upstream flow restriction such as a control valve, pressure
valve tailpipe connection to the test header, and various safety valve (PSV), or restriction orifice that passes a high mass
mitigations reduced dynamic stresses by 8-52% depending on flow of process fluid with a large pressure drop. In some cases,
the piping and type of mitigation. particularly with thin walled, large diameter pipe, the resulting
Acoustic and structural finite element models were vibration levels are severe enough to cause high-cycle fatigue
analyzed in order to determine the coincident modes that match damage and failure, usually at welded supports or branch
in both axial/circumferential shape and natural frequency and connections. Due to the high response frequencies (typically
compare coincident frequencies with measure stresses. The 100 - 3,000+ Hz), fatigue failures in some severe cases have
results show that observed peak stress frequencies do not been reported to occur within as little as several minutes of
generally correlate well with predicted coincident modes, and operation. The severity of AIV and likelihood of AIV-induced
that flow-induced turbulence excites frequencies below piping failures are complex physical phenomena that depend on
shell modes that can also result in significant stresses that multiple factors including an accurate characterization of the
combine with AIV. noise source(s), internal multimodal piping acoustics and
The width, thickness, and spacing of the stiffening rings all The IEC predicted noise for the series combination of the
affect the vibrational response of the header pipe and the control valve and PSV is 192 dB assuming a conservative
resulting dynamic strain at the tailpipe weld. Similarly, the liquid pressure recovery factor FL of 1 for both valves. A more
geometry of the tube bundles, while reducing low-frequency realistic FL estimate of 0.6 for the PSV produces a predicted
acoustic modes in the header, will set up additional high level of 187 dB.
frequency modes that may have not been present in the baseline The measured and predicted noise levels for several tests
case. Introducing tubes into the mean flow of the header will are shown in Table 2, where IEC (a) is computed with a PSV FL
also result in additional flow restriction and increased of 0.6, and IEC (b) is computed with a PSV FL of 1.0. The EI
turbulence generation. calculation is shown both in terms of power level, as the
standard produces, and converted to pressure by the method
MEASUREMENT RESULTS II: DYNAMIC PRESSURE described by Evans12, shown in Equation 2. The 6 dB sonic
Physical testing was performed at the valve manufacturer’s flow factor was included for the PSV for all EI calculations.
blowdown test rig in an attempt to replicate the scenario that The noise is observed to track both the flow and pressure ratio
led to valve d. An overview of several valve noise predictive as shown in Figure 7.
methods and measurement results from a test configuration
1
similar to this study have been presented by Evans 12. The 𝐿𝑊,𝑝𝑖𝑝𝑒 = 𝐿𝑃 − 10 log10 ( ) (2)
𝜋𝑎2
predicted valve noise is best described by International
Electrotechnical Commission (IEC) standard13, based on Comparing noise measurements from different locations
several parameters which define a noise generation regime and provides an assessment of the degree of decay with distance.
associated acoustic efficiency. The standard provides a sound The distance between ring A and B was 100” for all tests; the
power W which can be used to produce the predicted sound total straight line distance between ring B and C varied
pressure level as shown in Equation 1, where 𝜌 and 𝑐 are depending on the header pipe diameter, where the horizontal
evaluated downstream of the valve: distance from ring B to the center line of the elbow was 31” for
all tests. The distance from the center line of the elbow to
3.2𝑥109 𝑊𝜌𝑐
𝐿𝑃 = 10 log10 ( ) (1) ring C was 16”, 32”, and 41.5” for the 36”, 20”, and 12” header
𝐷𝑖2
tests, respectively, so measurements between rings B and C
compare results from non-uniform spacing.
The average RMS pressure level measured at the ring A
locations (12.75” from PSV throat) was 193.9 dB ref. 20µPa, which
is consistent with past measurements at similar conditions12.
The measured spectra at positions 1 (ring A), 5 (ring B), and 12
(ring C) for a representative case are shown in Figure 10, where
the response is broadband up to approximately 1 kHz, after
which the modal response in the tailpipe begins to contribute.
These spectra indicate a slight increase in noise from 300 –