Ranking of Compressor Station Noise Sources Using Sound Intensity Techniques

Transcription

1 THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS 345 E. 47 St., New York, N.Y GT-240 The Society shall not be responsible for statements or opinions advanced in papers or in discussion at meetings of the Society or of its Divisions or Sections, or printed in its publications. Discussion is printed only if the paper is published In an ASME Journal. Papers are available from ASME for fifteen months after the meeting. Printed in USA. Copyright 1987 by ASME Ranking of Compressor Station Noise Sources Using Sound Intensity Techniques W. D. JOHNS Senior Engineer, Station Design R. H. PORTER Senior Engineer, Station Design TransCanada PipeLines Toronto, Ontario, Canada ABSTRACT QA directivity of a source without reflective surfaces, dimensionless Local residential development and the introduction of more restrictive noise regulations in Canada and the United States are creating a need to improve the noise abatement systems at many existing industrial sites including pipeline compressor stations. The initial phase of any silencing program should include a study to identify and rank the noise sources. Until recently, this type of noise study has been qualitative and inexact, requiring a trial and error approach which addressed only one or two sources at a time and often resulted in a prolonged and costly silencing program. The use of sound intensity techniques to determine sound power levels of all noise sources results in lower costs, improved job scheduling and greater likelihood of success of a silencing program. This paper discusses a case study which uses sound intensity techniques to rank noise sources at a natural gas compressor plant powered by a gas turbine. NOMENCLATURE A e = excess attenuation caused by environmental conditions, db DI e = directivity index of the source in the direction e, db (Note: even if the source is nondirectie, DI = 3 db for hemispherical radiation) L = sound pressure level at a receiver located in 5 Pe the direction e, at a distance r, db re 2 x 10 N/m 2 L w sound power level of the source, db re watt NI low pressure gas turbine compressor rotational speed, rpm N2 high pressure gas turbine compressor rotational speed, rpm N3 driven equipment (gas compressor) rotational speed, rpm Q s spatial directivity, accounting for reflective surfaces, dimensionless 043 = total directivity factor, dimensionless r = distance of receiver from source, m Subscripts e = angle from a reference point, degrees INTRODUCTION To transport natural gas from the border of Saskatchewan and Alberta to distributors as far east as Montreal, Quebec, TransCanada PipeLines operates over 1,000 megawatts of compression power. Of this power, 83 is provided by gas turbines. Although the compressor stations along the pipeline were located remote from residential areas in the late 1950's and early 1960's, subsequent residential development has resulted in a need for noise abatement programs at some stations. Increased public sensitivity to sound and the reed for compression power additions have aggravated the situation. Legislators, becoming more aware of noise as an environmental issue, have introduced property line noise limits which are often below the existing site values. Noise control equipment, at the source or in the path to the receiver, is expensive to install and operate and often reduces the efficiency of a gas turbine compressor plant by introducing higher pressure drops in gas turbine flow paths. Therefore, it is important to optimize the silencing equipment to meet the far field noise requirements with minimum cost. Identification and ranking of the noise sources is essential for both new and existing installations. Only the sources which are contributing to the excessive noise levels need to be treated. Presented at the Gas Turbine Conference and Exhibition, Anaheim, California May 31-June 4, 1987

2 Frequently, a trial and error approach is used to attenuate compressor plant noise. Dominant sources are identified from far field sound pressure measurements by comparing far field noise spectra with near field spectra of probable sources. It is very difficult, however, to distinguish between spectra of sources when many sources exist in the near field. It is also difficult to know how much to silence a source and to know whether all the important sources are identified. Often a major source is treated, lowering the near field noise but reducing the far field levels only marginally because other sources start to dominate. Suppliers of components for compressor plants can rarely provide accurate unsilenced sound power levels for their equipment. As a result, compressor plants will often fail to meet the specified noise limits. Due to tight construction schedules, it is usually impossible to return any equipment for modifications and in these cases, expensive and sometimes unsatisfactory field retrofitting takes place. It is important that suppliers provide suitable noise data on their products and that the facilities be acoustically designed at the out set. The use of sound intensity techniques will provide better sound power information by determining sound power levels of individual sources without subjecting bulky equipment to the confines of anechoic or reverberant chambers. These sound power levels can be calculated from intensity measurements taken in situ in the presence of many sources. Using sound powers and correcting for directivity, distance, and excess environmental attenuation, a mathematical model can be generated to determine the effect of the major sources on far field sound pressure levels. This model allows the ranking of the sources in order of importance and provides a means to predict the impact of a noise abatement program. The results will be used to predict sources at other similar plants, and to provide information to suppliers to enable them to improve gas turbine package design. SOUND INTENSITY Sound power can be reliably calculated from sound pressure levels in a controlled environment, or in the free field where sources do not interfere with one another. If ambient noise levels are high and the sound field is reactive, however, only sound intensity measurements will enable calculation of accurate sound power levels. Sound intensity is the sound energy flux, a vector quantity describing the magnitude and direction of the net flow of acoustic energy. Therefore, the dimensions commonly used for sound intensity are W/m 2. By taking 10 times the logarithm of the i ratio of sound intensity to a reference value (10 Watt/m2), the sound intensity level can be expressed in decibels. The integral of the sound intensity over a surface is the sound power passing through the surface. The sound intensity and sound power levels can be expressed in terms of octave bands, third octave bands, or overall noise levels over any frequency range. There are different instrumentation packages on the market that can measure sound intensity levels. The instrumentation generally consists of a pair of microphones in conjunction with either a dual channel Fast Fourier Transform (FFT) signal analyzer such as the Bruel & Kjaer (B&K) type 2032 analyzer used in this case study, or a real time type sound intensity analyzer like the B&K For continuous level noise sources such as gas turbines, both types of analyzer will give similar results. For the study of an unsteady source such as a jack hammer, a real time instrument should be used to capture the peak levels. Sound intensity techniques have a number of inherent limitations associated with them: bias errors, resulting from the finite pressure difference approximation for particle velocity; phase mismatch errors due to phase differences in the microphones and analyzer channels; and reactivity errors resulting from phase mismatch of both the equipment and the measurement surface. The bias errors limit the accuracy in the higher frequencies while the phase mismatch errors limit the lower frequency capabilities. Reactivity errors could result at any frequency depending on the location and sound power levels of extraneous sources and the distance between the microphones. The microphone spacing should be selected correctly for the frequency range of interest in order to minimize the amount of error. These errors are fully discussed in the literature (Brock, 1985; Bruel & Kjaer, 1983; Rasmussen, 1985). In addition to sound power levels, an important factor in determining the effect of a source on the far field is the directivity of the source. There are two components to the directivity which can be described as directivity factors: the directivity which a source would exhibit if it was operating in an anechoic chamber or in the air without any reflective surfaces (Q A ); and the directivity effect on a source due to reflective surfaces, which can be termed spatial directivity (Q s ). Such sources as exhaust ducts, air intake ducts and vents radiate sound non-uniformly even if there are no reflective surfaces. The spatial directivity factor accounts for reflections from such items as the ground and walls. The spatial directivity factors for spherical, hemispherical and quarter spherical propagation are one, two, and four respectively. The total directivity factor Q9 is defined as the product of QA and Q s, and this total directivity factor is translated into a directivity index (DI e ), expressed in decibels, by the following equation: DI e = 10 Log Q e 10 Log Q A 10 Log Qs MEASUREMENT TECHNIQUES Sound intensity measurements for any particular source will use one of three types of control surface: conformal, i.e. conforming to the shape of the object; hemisphere, i.e. a hemispherical "cover" placed over the source; box, i.e. a box-shaped "cover" placed over the source; or a combination of these surfaces. The control surfaces are determined using coordinates relative to the object of interest. The physical size of the majority of the sources examined in the case study dictated the use of a box technique. This technique is best explained by describing sound intensity measurements over one of the sources investigated - an inertial air filter extraction fan. A box shape was constructed over the fan as shown in Figure 1 and the area of each of the five open sides was determined. The sixth side of the box was covered by the steel plate of the filter house. Since the fan and duct did not radiate sound uniformly through each (1 ) 2

3 side of the box, the sound intensity needed to be measured for each of the five sides. Before taking readings of the average sound intensity for the box, the choice was made between the use of a "arid" or a "sweeping" technique. The grid technique involves constructing a real or imaginary grid of equal area shapes over a surface and taking sound intensity measurements at the centres of each of these shapes. The grid size should be small enough that the intensity does not vary greatly throughout the shape. An identical grid is set up within the sound intensity computer program (B&K WW9078) and measurements are then taken systematically. The operator ensures that the probe is in the correct location in the centre of the shape and perpendicular to the box surface. Once all measurement points are stored, the computer program will display the sound intensity results in tabular or graphical form. The sound power for each of the grid areas as well as the total sound power of the complete surface is calculated by the computer. The sweeping technique is a space averaging process taken across a complete side. The probe is kept at right angles to the surface and swept uniformly across the surface while the analyzer averages the sound intensity. The sweeping technique was used for each of the five sides of the imaginary box over the extraction fan, using 250 averages over each of the sides or sectors. The results are described in the case study section. CHOICE OF FREQUENCY RANGE As described in the sound intensity section, the frequency range over which measurements are taken results in various values of bias, phase mismatch, and reactivity errors. These errors can be minimized by selection of the proper combination of frequency range and space between the probe microphones. The frequency range of 56 Hz to 5600 Hz chosen for the case study allowed the use of a 12 mm spacer for all measurements and resulted in reasonable accuracy. As shown in Figures 2 and 3, the Hz range adequately covered the frequencies having the greatest contribution to sound pressure levels at the far field reference point 45 m to the northeast. This range also covered only two frequency decades, a span which could be handled in a timely fashion by the intensity analyzer. CASE STUDY Site Description The gas turbine powered compressor plant chosen for the noise source ranking study does not suffer from noise complaints at this time, but it is an installation where equipment did not meet specified noise levels upon installation and one near which residential development may soon occur. Sound levels at 100 metres from the compressor building vary from 58 to 68 dba depending on direction and operating conditions. The compressor package consists of an axial inlet centrifugal compressor driven by a 12 MW gas turbine. This type of package is representative of the second generation aero derivative gas turbines at other locations on TransCanada's system, and the installation is therefore considered typical for the purpose of noise source identification. The compressor building is shown in Figures 4 and 5 where the two most often identified gas turbine noise sources, the air intake and the exhaust can be clearly seen along with a number of other sources which were investigated. The design of this particular plant included silencers on the air intake and exhaust of the gas turbine, an acoustic enclosure over the gas turbine, and acoustic insulation over the above ground gas piping. Since the major sources within the building were believed to be adequately treated, the compressor building was not designed with acoustics in mind. The equipment specifications stated that the compressor building should be considered acoustically transparent. Summary of Measurement Techniques A variety of measurement techniques were used to calculate power levels of the different sources. The techniques used and the pertinent sources are listed in Table 1. The instrumentation used and the survey conditions are summarized in Appendix A. TABLE 1 SUMMARY OF MEASUREMENT TECHNIQUES Control Surface and Technique Source Box/grid and sweep Box/sweep Conformal/sweep Exhaust plume Exhaust silencer walls Gas turbine vent Extraction fan Exhaust duct Air intake screen Air intake plenum Exhaust plenum Intake filter house Louvres Walls Mandoor Blowout panels Overhead door Oil cooler Measurement Results The computer program (B&K WW9078) used to determine the sound power levels displays the measured intensity and the area of each surface. As shown in Table 2 prepared for the extraction fan, the program also computes the sound power of each side of the imaginary box, the total sound power from all sides, and, if desired, lists the sound power level versus frequency as either a table or histogram (Figure 6). Sector 1 top 2 side 3 side 4 bottom 5 end TOTAL: TABLE 2 EXTRACTION FAN SOUND POWER Intensity (dba) 88.7 dba dba dba dba dba dba+ Area ( m 2) Sound Power (dba) 91.4 dba dba dba dba dba dba+ 3

4 The total sound power of the extraction fan is dba, which is the sum of the sound powers of all five sides of the box. Since the far field criterion is expressed in dba, all the intensity and sound power levels have been 'A' weighted. Various sources around the compressor plant were investigated by applying the sound intensity measurement techniques, and sound power levels of the sources were determined. The power levels are used as a basis for determining which sources have the greatest effect at any given location. Source Ranking and Discussion Sources are listed in Table 3 in order from highest sound power level to lowest sound power level. A more meaningful ranking, however, takes into account the placement and orientation of the sources relative to receivers in the far field. For example, if residences are located to the east of the compressor plant, sound from only some of the sources will propagate in that direction, and the directivity factors for the pertinent sources must be taken into account. The sound pressure level attributable to a source, in a specific direction and at a particular distance is calculated from the following wave divergence formula from Beranek (1971). Lpe = L + DI e - 20 log r - A e - 10 log 47 A point 45 metres to the northeast of the compressor building was chosen as a reference point for comparison of calculated and measured sound pressure levels. To make this comparison, a computer model has been prepared. Compressor plant sound sources are viewed in perspective from the reference point (Figure 4) and correction factors for distance and directivity are applied to the source power levels for each third octave band from 63 Hz to 5000 Hz. The directivity data used for the sources is summarized in Table 4. The sound pressure levels are entered into a spread sheet type program in which the pressure levels are added together in each third octave and the overall levels for all sources are determined. The calculated total at 45 metres due to the top nine sources affecting the northeast location is then compared to the sound level measured at the same location. The output of the computer model is shown in Table 5, and it can be seen that the calculated and measured values of overall sound level are identical. There are differences in the calculated and measured third octave bands, but the average difference is approximately 3 db. The 10 db difference in the 4000 Hz band could have been lower, in fact, had it not been for a gas turbine speed change. The source responsible for the 63 clea peak in the 4000 Hz third octave band in the far field measured data is related to the power turbine speed, N3. The narrow band spectrum at 45 m northeast (Figure 3) shows this peak to be 400 Hz wide at the skirt and centred on the blade pass frequency of the power turbine (N3 = 5240 RPM, Number of blades = 53). The power turbine speed changed from 5240 RPM when the far field measurements were taken to 6000 RPM when the exhaust support and duct area were studied. This change in speed shifted the peak from 4620 Hz to 5300 Hz. Although the peak remained in the 5000 Hz third octave band, the broad skirt moved out of the 4000 Hz third octave band to totally within the 5000 Hz band. Vithout this shift in the peak, the IC db difference in the model at 4000 Hz would be lower. Some of the differences between predicted sound pressure levels and measured sound levels in other third octaves are also caused by variations in unit speed throughout the course of the measurements. The noise due to the power turbine that is observed in the exhaust support area is believed to be caused by a slip joint in the exhaust duct. To allow for thermal expansion, the slip joint consists of an inner and an outer sleeve which results in a sizable air gap. The predicted levels at 45 m rely on directivity factors for the sources. Although references are available for this information, the data is in a generalized form and can be difficult to apply. For instance, the exhaust silencer was modelled as a vertical stack with an outlet of 2600 mm X 2600 mm which fits the available references. The silencer in fact has two full splitters and two half splitters extending to the outlet with the splitters running in an east/west direction. The outlet therefore consisted of three rectangular areas with probably a different directivity in the east and west compared to the north and south. The length of the splitters would affect directivity as well by reducing the number of cross modes. The directivity would also be affected by the length of the splitters. Since no reference has been found to account for this geometry, the 6 db overprediction at 250 Hz is assumed to be partially due to directivity of the exhaust plume. Time constraints prohibited continuous monitoring of reactivity during measurements, and it is possible that reactivity errors occurred at some frequencies. Modified software, available for the B&K 3360 real time analyzer but not currently available for the B&K 2032 used for this study permits routine monitoring of reactivity in the third octave bands, and this monitoring is important. Reactivity errors may have contributed to underestimation of intensity in the 63 Hz third octave band and possibly in the 80 Hz third octave band. By examining Table 5, the sources contributing most to each third octave band can be found and ranked in order of their contribution to the overall sound pressure level at the reference point. The sources are ranked as follows: gas turbine vent, exhaust duct (support area), extraction fans, exhaust plume tied with the building louvres (considering the fact that an over-estimate may have occurred for the exhaust particularly in the 200 to 315 Hz bands), filter house tied with the exhaust silencer walls, exhaust plenum walls, and air intake screen. This order is quite different from that shown in Table 3 where the sources were ranked by overall sound power levels only. The directivity indices which were used in calculating the sound pressure levels at the reference point have a major effect on source ranking. By applying silencers to the sources in the computer model, the effect of reducing the sound power level of any one source or group of sources can be observed as shown in Table 6. For example, application of insertion loss data for a typical commercial silencer will reduce the power level of the gas turbine vent by 24 db, but the calculated total sound pressure level from all sources will only be reduced by 3 db in the northeast. If the exhaust duct is acoustically treated instead, the sound pressure level will be reduced by 1 db, but if both the vent and the exhaust duct are 4

5 silenced, the combined effect in the far field will be a reduction of 5 db. When the cost of equipment is combined with the predictions of the computer model, the optimum solution can be found for any given far field noise requirement. For example, to achieve at least a 4 db reduction in the northeast, modifications can be made to either the combination of the gas turbine vent and the exhaust duct or the gas turbine vent and the extraction fans. The optimization process should include a study of the sensitivity of cost to silencer insertion loss and equipment performance, following which the lowest cost alternative can be implemented. CONCLUSIONS AND RECOMMENDATIONS As discussed, the top five sources causing high sound pressure levels in the northeast direction are the gas turbine vent, the exhaust duct, the air intake filter extraction fans, the exhaust plume, and the building louvres. Although the computer model contains some inaccuracies because of field reactivity, speed changes and the selected directivities, these factors will not seriously affect a noise abatement program. The items to be silenced are known and the required silencing can be determined within reasonable tolerances consistent with the tolerances on standard silencing equipment. Further work is required in standardizing sound intensity measurement techniques in order to control the methods used by suppliers in obtaining sound power data. Both the American National Standards Institute (ANSI) and the International Organization For Standardization (ISO) are preparing such standards. The gas turbine and the natural gas industries, both users and suppliers, need to lend their support to the development. The directivity factors for sources can create significant errors. Additional work in this area is required, particularly with gas turbine exhaust. Sound intensity techniques, along with artificial noise sources could be used in some cases to determine directivities for existing facilities. This information would be useful for future designs. ACKNOWLEDGEMENTS The authors wish to thank Bruel & Kjaer Canada for their technical advice. The authors also wish to acknowledge Mr. Ian Wilson, Undergraduate, Applied Mathematics, University of Waterloo, for his assistance with data acquisition and reduction. REFERENCES American Gas Association, 1969, Noise Control for Reciprocating and Turbine Engines Driven by Natural Gas and Liquid Fuel, American Gas Association, New York, Table 34. Beranek, Leo L., 1971, Chapter 7, Noise and Vibration Control, McGraw-Hill Book Company, page 165. Brock, Michael, 1985, "Intensity Measurement Using a Tape Recorder", Intensity Measurements, The Analysis Technique of the Nineties, Bruel & Kjaer Publication BA (English), pages Ll - L5. Bruel & Kjaer, Sound Intensity Analysing System Type 3360, Bruel & Kjaer, Naerum, Denmark, pages 3-5. Burgess-Manning, Industrial Silencing - Vents and Blowdowns, Burgess-Manning Bulletin , Table 4. Rasmussen, Per, 1985, "Phase Errors in Intensity Measurements", Intensity Measurements, The Analysis Technique of the Nineties, Bruel & Kjaer Publication BA (English), pages B6 - B7. Source TABLE 3 SOUND POWER LEVELS OF SOURCES Power Level (dba) Exhaust Plume 114 Gas Turbine Vent 113 Air Intake Filter Extraction Fans (Two) 108 Exhaust Duct 107 Building, comprising 105 East Wall, consisting of 101 Louvres 101 Mandoor 77 Blowout Panels 79 Remainder of Wall 80 West Wall, consisting of 102 Louvre 92 Overhead Door 82 Wall section around piping 97 Mandoor 77 Blowout Panels 82 Oil Cooler 97 Remainder of Wall 80 North Wall, consisting of 84 Blowout Panels 70 Mandoor 77 Remainder of Wall 81 South Wall, consisting of 94 Louvre 92 Overhead Door 82 Blowout Panels 88 Remainder of Wail 79 Roof (calculated using louvre 99 & wall data) Ventilators 99 Remainder of Roof 89 Air Intake Screen 102 Exhaust Silencer Wall 100 Air intake Filter House 99 Exhaust Plenum 94 Air Intake Plenum 90 5

6 TABLE 4 DIRECTIVITY INDICES USED FOR MAJOR SOURCES Freq. (Hz) Directivity Indices for the Major Sources (db) [Spatial Directivity Factor MO for Reflective Surfaces Shown in Brackets] Exhaust 1 Exhaust Exhaust Exhaust Gas Turbine Supports Silencer Walls Plenum Walls Vent 2 Extraction Fan 2 Air Intake Screen 1 Air Intake Building Filter House Louvres (4) (4) (4) V American Gas Association, 1969, corrected for spatial directivity. 2 Burgess - Manning Bulletin corrected for spatial directivity. TABLE 5 SUMMARY OF CONTRIBUTIONS OF MAJOR SOURCES TO FAR FIELD LEVELS Third Octave Frequency (Hz) (1) Exhaust Exhaust Supports (3) Exhaust Silencer Walls (4) Exhaust Plenum Walls (5) Gas Turbine Vent Sound Pressure Level (dba) at 45 m Northeast (6) Extraction Fans (7) Air Intake Screen Air Intake Filter House (9) Building Louvres (East) (10) Total (Sources 1-9) (II) Measured Total (12) Difference (11-10) TOTAL Operating Conditions NI (RPM) N3 (RPM) Power (MW)

7 ... [,... TABLE 6 SILENCING BENEFITS Source(s) Treated Reduction in Source Power Level (dba) Reduction in Overall Sound Pressure Level in the Northeast Direction (dba) Gas Turbine Vent (A) 24 3 Exhaust Duct (B) 23 1 G.T. Vent & Exhaust Duct (A + B) - 5 Extraction Fans (C) 22 0 G.T. Vent & Extr. Fans (A + C) - 4 Exhaust Duct & Extr. Fans (B + C) - 2 G.T. Vent & Exhaust Duct & Extr. Fans (A + B + C) - 8 Exhaust Plume (D) 9 0 (A + B + C + D) 10 Building Louvres (E) 16 0 (A +B+C+D+ E) Is' I,... Figure 1 - Extraction Fan Control Surface 7

8 k 4k 6k Frequency (Hz) 8k 10k 12k Figure 2 - Far Field Noise Spectrum Hz WINFIU MI " 11=11F ur.glrlwr lk 2k 3k 4k 5k 6k Frequency (Hz) Figure 3 - Far Field Noise Spectrum 0400 Hz 8

9 Figure 4 - Compressor Building Viewed from the Northeast (45 m from air intake) Figure 5 - Compressor Building Viewed from the Southwest 9

10 cl 100 'Cs > Figure 6 - Extraction Fan Sound Power APPENDIX A INSTRUMENTATION AND SURVEY CONDITIONS Instrumentation: Bruel & Kjaer Dual Channel Signal Analyser Type 2032 Bruel & Kjaer Sound Intensity Probe Type 3519 Bruel & Kjaer Microphone Set Type mm microphone spacer Foam windscreen Bruel & Kjaer Microphone Power Supply Type 2084 Bruel & Kjaer WW9078 Sound Intensity Software Package Bruel & Kjaer Pistonphone Type 4220 Hewlett Packard 9816 Computer Hewlett Packard 9122 Dual Disc Drive Hewlett Packard Thinkjet Printer Hewlett Packard 7470A Plotter General Weather Conditions (over measurement period of 4 days) Wind - variable wind conditions, prevailing from NW Temperature - 10 C to 20 C Average Pressure mb Precipitation - none during measurements Operating Conditions (over measurement period of 4 days) Gas Turbine Power MW to 12.1 MW Low Pressure Axial Compressor Speed (NI) RPM to 8090 RPM High Pressure Axial Compressor Speed (N2) RPM to RPM Gas Compressor Speed (N3) RPM to 6030 RPM 10