Challenges in delivering effective noise management at gas compres- sor stations Carl Christian Hantschk 1 and Marco Geisler 2 Müller-BBM GmbH Helmut-A.-Mueller-Strasse 1 - 5 82152 Planegg GERMANY

ABSTRACT The basic principles applied in gas compressor stations to manage noise and demonstrate compliance with statutory acoustic requirements are not different from those used with most other industrial activities. Well-known and proven standard procedures involve sound emission assessment, acoustic modeling, model verification, development of a noise control concept and its optimisation until com- pliance can be demonstrated. However, noise control for some of the typical equipment in gas com- pression is extremely demanding in terms of both acoustic and operational/safety aspects so that it requires special attention. Examples are the exhausts of the gas turbines driving the compressors and the extensive network of gas piping. Because of their high and strongly varying sound emissions the exhausts pose a significant challenge in designing effective silencers. In addition, flow-generated self-noise in the silencers is a risk that needs to be avoided. For the piping, sound attenuation by acoustic lagging is in conflict with corrosion and maintenance/accessibility problems and makes an optimised lagging concept an ambitious task. The present contribution illustrates typical problems encountered and possibilities to deal with them. It is based on data and experiences with gas com- pressor stations with very stringent acoustic requirements.

1. INTRODUCTION

In most countries industrial activities have to comply, among other things, with statutory acoustic requirements. Typically, these requirements are defined in the plant's license to operate (e.g. a Pollu- tion Prevention Control Permit) as maximum permissible noise levels (noise limits) at specific loca- tions in the neighborhood of the plant and/or at workplaces inside of it (Noise Sensitive Receptors, NSR). Such acoustic requirements are met if the prevailing noise levels at these NSRs, for all relevant operating conditions of the plant, do not exceed the corresponding noise limit values.

For existing installations, the quickest and easiest way to demonstrate that this is the case is by direct sound level measurements at the NSRs in the relevant operational modes of the plant. However, rather often this may not be possible for different reasons, for example: • Valid measurements cannot be taken because the plant cannot be operated in the conditions rele- vant for the noise limits.

1 Carl-Christian.Hantschk@mbbm.com

2 Marco.Geisler@mbbm.com

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW

• The situation at the NSRs does not allow to determine the sound pressure level received from the plant – the specific sound level – by measurements with sufficient accuracy (e.g. if the residual sound is too high). • The NSRs are not accessible for measurements. Another situation where direct measurements at the NSRs are of little help is if the specific sound levels obtained from the measurements are too high, i.e. they show non-compliance. This is because the measurement result at the NSR does not provide any clues as to how compliance can best be achieved.

In such situations an alternative approach is required. 2. SOUND PROPAGATION MODELING

A well-known and proven method in situations as outlined above is to determine the specific sound level at the NSRs by performing a sound propagation calculation [1]. In this case the following typical procedure can be used to demonstrate compliance: a) Assess the sound emission characteristics (typically power level and directivity of the emitted

sound) for all relevant individual sound sources and sound transmission paths of the plant by measurements on site with these emitters operated in the mode relevant for the statutory acoustic requirements. Here, the term "sound source" is used for items that directly generate and radiate sound whereas "sound transmission paths" are items that do not generate sound directly by them- selves, but are transmitting sound generated by other sources. An example for the latter is the exhaust stack outlet opening that, ultimately, radiates part of the sound generated in the gas tur- bine. b) Measure the ambient noise level at a suitable "checking location": as far away from the plant as

possible, but close enough to have the prevailing ambient noise level dominated by the sound emissions from the plant. During the measurements the plant must be run in a stable and defined condition. This data will be used for quality assurance in step d). c) Set up a three-dimensional acoustic calculation model (acoustic model) of the plant, comprising

all relevant sound sources and sound transmission paths in the plant and their respective sound emissions as determined from the measurements from a). In addition, all other items that are relevant from an acoustics point-of-view need to be taken into account in the model – for example topography and buildings and other obstacles that can have an influence on propagating sound by acting as barriers to sound waves and/or by reflecting sound waves. d) Perform a sound propagation calculation with the model for the operating condition from step b).

Compare the calculated noise level at the checking location to the level obtained from the meas- urements to verify that the model correctly predicts the sound field around the plant. e) Use the verified model to calculate the sound pressure levels received from the plant at the NSRs

for the operating conditions relevant according to the statutory acoustic requirements. Before compliance is obtained in step e) – i.e. the sound pressure levels received from the plant at the NSRs do not exceed the maximum permissible levels – it will often be necessary to apply or upgrade noise mitigation measures on specific sound emitters in the plant. For an efficient noise control plan- ning the acoustic model is of great value. It can be used to identify the sound sources and sound transmission paths most relevant for the sound received at the NSRs and to evaluate different noise control options. The final objective is to put a noise control concept in place that, after going through the applicable steps of the procedure described above, achieves compliance in step e) but also mini- mises impacts on non-acoustical aspects such as safety, performance, efficiency, maintenance, costs etc.

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW

For plants still in the planning stage, where no measurements can be taken at all, an initial run of the procedure must be made replacing the measurement data from a) by data from predictive calcu- lations or databases. In this case, no verification of the model (see steps b) and d) above) can be performed. Once the station is operational the process workflow is as described.

Note that what has been outlined here in the context of demonstrating compliance may need to be repeated in part at regular intervals within a proper noise management to ensure that equipment sound emissions or the effect of noise control measures have not changed and, in particular, when acousti- cally relevant modifications or additions to the plant are made.

3. GAS COMPRESSOR STATIONS

In typical gas compressor stations natural gas is compressed from a lower pressure level in the in- coming pipeline to a higher pressure level and fed into the outgoing pipeline. This is mainly done to improve efficiency in transporting the gas in the pipeline system (basically by increasing the mass flow rate) and to compensate for pressure losses incurred when the gas moves through the pipework over long distances. The core equipment of a compressor station are the gas compressor units. In the stations used as examples in this paper, these units consist of the actual centrifugal gas compressor, a power gas turbine that drives the centrifugal compressor and a gas generator turbine that drives the power gas turbine. The exhaust gases from the power gas turbine are emitted into the open via exhaust stacks.

When the procedure described in section 2 is applied in noise management for these gas compres- sor stations, some of the typical equipment involved requires special attention. In the following, focus is set on two important examples for such equipment – the exhausts of the gas turbines driving the compressors and the extensive network of gas piping in the station.

4. POWER GAS TURBINE EXHAUST STACKS

4.1. Sound emissions and silencer design The sound emitted from the exhaust stack openings of the power gas turbines has very high sound power levels and the openings are usually located at an elevated position. It is, therefore, almost always the most important contributor to the sound pressure level received from a compressor station at its NSRs in the neighbouring surroundings. To keep this contribution within acceptable limits (i.e. compliant with the statutory acoustic requirements), silencers are needed for the stacks. Once the sound emission characteristics of the stack exit opening are known, the modeling procedure described in section 2 can be applied in the design process of an effective silencer by determining whether the acoustic performance of a specific design is suitable to ensure compliance.

However, determining the sound emission characteristics of the stack exit opening poses a signif- icant challenge: these emissions vary strongly because they depend on the operational state of the power gas turbine which, on the other hand, is dictated by the varying parameters of the compression process. Depending on the upstream and downstream gas pressure and the required gas flow rate, the compression process' power demand changes and many operational parameters of the gas turbine need to be set accordingly to satisfy it. Examples for these parameters are rotational speed, inlet guid- ing vane position, bleed valve setting, by-pass ratio and the actual firing rate.

The three diagrams in Figure 1 show the total A-weighted power level PWL of the sound emitted from the exhaust stack openings of different power turbines (left: Solar Mars 100, middle: Siemens SGT-100 twin-shaft Tornado, right: Solar Taurus 70S) and exhaust system arrangements (each in- cluding one or more silencers). Data has been collected in measurements within the last 13 years with

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW

the operational state of the turbines at arbitrary points in time, i.e. as necessary for the gas compres- sion demand at that time. The left diagram shows data for four different units of identical design, the middle and right one for three units, respectively. The PWL is plotted versus compression work done on the gas stream per unit of time P compr , calculated for the respective pressure increase upstream and downstream of the centrifugal compressor, the gas mass flow rate and the simplifying assumption of adiabatic compression.

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW

Figure 1: Total A-weighted power level PWL of the sound emitted from the exhaust stack openings of different power turbines and exhaust system arrangements versus compression power.

A comparison between the three diagrams – i.e. between different power turbine types – is of little meaning because the exhaust systems and the silencers fitted are different for each of the three instal- lations. However, within each diagram (= for three or four units of identical design), a significant spread in the PWL from the measurements can be seen of up to about 15 dB, 7 dB and 17 dB, respec- tively. There seems to be a rudimentary correlation between P compr and PWL, but the spread in the data indicates that P compr is a quantity too general, failing to take into account effects and operational parameters that are relevant in driving the turbine's sound emissions.

Figure 1 suggests that – in lack of a suitable calculation procedure for PWL from the operational parameters – a sufficient number of measurements could be used to determine the maximum sound emission that must be expected from the exhaust stack opening within the long-term operational range of the system. This information could then be used to design an effective silencer for the exhaust system. Unfortunately, this is of little practical value because it will normally not be possible to run a unit without silencer to take such measurements nor can measurements be taken during the planning phase of a compressor station.

An alternative is to rely on the power turbine manufacturer's sound emission data, but using ade- quate caution: because of the wide range of possible operating scenarios, supplier data is often limited to only few representative modes, for example full load and a generic part load, both at fixed speeds. As the data reported in Figure 1 and also information from manufacturers (e.g. [2]) indicate, the sound emissions in these modes may be lower than the maximum emissions relevant for the acoustic re- quirements at the NSRs. Accordingly, for designing effective silencers for the exhaust system, a safety margin may need to be added to the sound emission data available from the manufacturer. The spread in PWL in Figure 1 can serve as a guideline for this margin but at the risk of being overly

‘Sound power level PWL in dB(A) ‘Solar Mars 100 Siemens SGT-100twin-shatt Tornado Soler Taurus 70S

conservative (i.e. of overestimating the actual sound emissions). To reduce this risk additional con- sultation with the engine manufacturer is recommended to check whether data is available for oper- ating conditions of the power turbine other than typical.

4.2. Flow-generated silencer self-noise A silencer installed in the exhaust system of the power turbine attenuates the sound from the upstream (turbine) side to a lower level downstream of the silencer. However, additional sound may still be generated downstream of the silencer if the exhaust gas flow passes deviations, restrictions and ob- stacles in the ductwork, such as transition pieces, bends, guiding vanes and dampers or the silencer baffles themselves (generally: flow resistances). These flow-generated sound contributions will leave the exhaust system without any attenuation by the silencer. Often this is not a problem, as the level of the sound originating from the turbine – even after having been attenuated by the silencer – will still be significantly higher than any sound generated by the flow downstream of the silencer. As a result, the latter is then irrelevant for the total sound emission from the exhaust system outlet opening.

However, this situation is often different in gas compressor stations where, typically, the flow velocities in the power turbine exhaust system – and, therefore, the levels of flow-generated sound – can be rather high. If a silencer reduces the sound from the turbine enough and/or if flow velocities are high enough, then sound generated by the downstream flow will make a noticeable contribution to the total sound emission from the exhaust outlet opening. In this case it becomes of vital importance to take flow-acoustic effects downstream of the silencer into account in the noise control planning and the design of the exhaust duct system. In particular, to make sure that the exhaust outlet noise limits targeted for are actually met on site, flow resistances of any kind should be avoided wherever possible or need to be acoustically optimised if located downstream of the silencer.

As an example, experiences from a gas compressor station where the power turbine exhaust stacks had been upgraded with high-performance silencers are reported in the following. In the acceptance testing during commissioning the stack just barely failed to meet the target noise level. The main reason were sound contributions marked by a distinctive peak in the 500 Hz and 400 Hz third-octave bands in the spectrum of the sound emitted from the stack outlet opening – as can be seen in the red curve in the diagram on the left in Figure 2.

Further testing indicated that the performance of the silencer was satisfactory and it was suspected that flow-noise generated at four lifting lugs downstream of the silencer was the source of the problem because of the high flow velocities of around v = 35 m/s. The lugs (see Figure 2, right) had been welded to the inside of the stack casing for reasons of shipping size restrictions. An estimative calcu- lation based on a Strouhal number of Sr = 0.21 and the lug sheet metal thickness of L = 15 mm yields a vortex shedding frequency at the lugs of f = Sr  v / L = 490 Hz which matches well the location of the peak in the spectrum.

Measurements with a 1:1 model of a lifting lug were performed in a flow-acoustics test stand and confirmed the suspected sound generation mechanism and also that closing the holes in the lifting lugs was a simple and effective countermeasure. After welding plugs flush into the lugs on site the sound emissions from the stack outlet opening did no longer show the peak in the spectrum – see the blue curve in the diagram on the left in Figure 2 – and the sound emissions, now reduced by about 2 dB, did comply with the acoustic requirements.

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW

Figure 2: Left: Averaged A-weighted sound pressure level spectrum SPL as measured at 1 m distance from the outlet opening of a power turbine exhaust stack with high-performance silencer installed, with open and closed lifting lug holes. Right: lifting lugs responsible for the peak in the 500/400 Hz bands (red curve).

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW

5. GAS PIPEWORK

A characteristic feature of a compressor station is its extensive network of gas piping that can com- prise several hundred meters of piping in the open. Compressing and moving the gas involves energy- intensive, high-speed centrifugal compressors that create a highly turbulent flow with significant gas velocities and many kinds of valves and accessories, as well as other flow resistances with pressure drops. All of these items and effects are associated with a corresponding mechanical and flow-me- chanical generation of sound which is partly transmitted via the pipe walls and radiated from the pipe surfaces to the outside as airborne sound. Because of the length of the piping and the large diameter of some of it, the surface area and, as a consequence, the total power level of the sound radiated from it may be quite significant.

It is therefore very important to include the gas piping into the sound propagation modeling pro- cedure described in section 2. This is often a time-consuming process, in particular because assessing the piping sound emissions (step a) in section 2) involves extensive measurements or calculatory predictions for all acoustically relevant parts and elements of the pipework. Nevertheless, only with the piping properly included the model will correctly predict the sound pressure levels received from the compressor station at the NSRs and thus can be applied for noise control planning. To be able to do so is of great importance because, although noise control as it is commonly used for piping is rather simple in principle – typically an acoustic insulation (lagging) consisting of one or more layers of porous (sound absorbing) material and a cladding made of sheet metal is applied to pipes with excessive sound emissions – determining an optimised lagging concept is an ambitious task. This is mainly for two reasons: • The sound emissions from different parts/sections of the pipework and their significance at the NSRs will be very different, depending on the intensity of the radiated sound, the area of the radiating surface and the location of the respective part. Note that a piece of pipe with lower

‘SPL at 1m in dB(A) “iting ug les — open closed 80. Frommeasurements: 70. 60. 50. 40. 30. —1—t 46 32 63 125 250 500 tk 2k 4k 8k 1 in Fz Liting lugs for silencer mounting

intensity of the radiated sound may be more important at an NSR than a piece with higher intensity if the first is very long and the second rather short. • Lagging can promote corrosion of the pipe and limits accessibility of the pipe for inspection and maintenance. Accordingly, the model with the pipework properly included is the decisive tool for determining an optimised lagging concept, i.e. a concept that achieves compliance with all acoustic requirements at the NSRs but, at the same time, reduces the length of pipe that is lagged to a minimum.

To illustrate the importance of noise control at the piping as such and the difference between a generic lagging concept and an optimised one, Figure 3 shows calculated sound pressure level con- tour plots in a compressor station for different lagging configurations for the station pipework.

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW

Figure 3 (partly reproduced from [3]): Calculated A-weighted sound pressure level contour plots for different lagging configurations for the station pipework (red lines). Top left: no lagging – non-com- pliance with acoustic requirements. Bottom left: standard acoustic lagging on 663 m of piping – acoustic requirements complied with. Bottom right: high-efficiency lagging applied to 346 m of pip- ing in total – acoustic requirements complied with.

Figure 3, top left shows the situation without any lagging applied to any of the pipework: it can be seen that high noise levels prevail with A-weighted sound pressure levels SPL reaching up to 96 dB(A) in some locations accessible to workers. In this state the total A-weighted sound pressure level received from the station at the closest inhabited NSR is 41.2 dB(A). The acoustic requirements of SPL ≤ 85 dB(A) at all workplaces and of SPL ≤ 40 dB(A) at the NSR are not complied with. The

bottom left and right in Figure 2 show the state with standard acoustic lagging and a generic lagging concept (left) and special lagging and an optimised concept (right) applied to the pipework in a way as to reach compliance with these requirements: for both configurations the A-weighted sound pres- sure level in the station is 85 dB(A) at maximum in workplaces and lower than 40 dB(A) at the closest inhabited NSR (left 39.2 dB(A), right 38.4 dB(A)). However, whereas the standard lagging with the generic concept shown in the bottom left plot involves 663 m of lagged piping, the special lagging with the optimised concept shown in the bottom right plot has only 346 m of lagging and even achieves a 0.8 dB lower noise level at the closest NSR. 6. KEY MESSAGE

Noise management for gas compressor stations relies on the same basic principles than most other industrial activities. However, especially if the acoustic requirements are stringent, noise control and demonstrating compliance can be a challenge and some of the typical equipment requires special attention. With focus on these challenges, the following is recommended for an effective noise man- agement in gas compression: • Set up a three-dimensional acoustic calculation model for sound propagation calculation of the compressor station and its surroundings and, if possible, calibrate it by measurements. • Include the gas piping in the station into the acoustic model. The total power level of the sound radiated from piping may be quite significant although the hearing impression may suggest oth- erwise because piping is a distributed sound source. • Use the model for an efficient and optimised noise control planning and for demonstrating and verifying compliance with the acoustic requirements. • Keep the model up to date, in particular, when acoustically relevant modifications or additions to the station are made. • Pay proper attention to the sound emissions from the exhausts of the power turbines driving the compressors. These emissions vary strongly depending on the operating conditions. As a basis for designing silencers for the exhausts either take sound emission measurements for a sufficient number of operational states or make sure that data from the manufacturer is for the relevant range of operational states. • Be aware that flow-acoustic effects downstream of a silencer for the turbine exhaust may be im- portant and take them into account in the (noise control) design. As a rule, flow resistances of any kind should be avoided in these parts of the exhaust system, if possible. 6. REFERENCES

1. BS 4142: Methods for rating and assessing industrial and commercial sound. 2014. 2. Solar Turbines Incorporated: Noise Prediction Guidelines for Industrial Gas Turbines. 2005. 3. Geisler, M.; Hantschk; C.-C.: Optimized noise control for gas compressor station pipework. Inter-

Noise 2016.

i, orn inter.noise 21-24 AUGUST SCOTTISH EVENT CAMPUS ? O? ? GLASGOW