A Different Approach to Gas Turbine Exhaust Silencing

Transcription

1 74-GT-26 Copyright 1974 by ASME $3.00 PER COPY $1.00 TO ASME MEMBERS 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 or Proceedings. Released for general publication upon presentation. Full credit should be given to ASME, the Professional Division, and the author (s). A Different Approach to Gas Turbine Exhaust Silencing MARV WEISS Technical Resources Operation, Gas Turbine Products Division, General Electric Co., Schenectady, N.Y. A unique method for silencing heavy-duty gas turbines is described. The Switchback exhaust silencer which utilizes no conventional parallel baffles has at operating conditions measured attenuation values from 20 db at 63 Hz to 45 db at higher frequencies. Acoustic testing and analyses at both ambient and operating conditions are discussed. Contributed by the Gas Turbine Division of The American Society of Mechanical Engineers for presentation at the Gas Turbine Conference & Products Show, Zurich, Switzerland, March 30-April 4, Manuscript received at ASME Headquarters November 13, Copies will be available until December 1, THE AMERICAN SOCIETY OF MECHANICAL ENGINEERS, UNITED ENGINEERING CENTER, 345 EAST 47th STREET, NEW YORK, N.Y

2 A Different Approach to Gas Turbine Exhaust Silencing MARV WEISS The most difficult and costly silencing requirement related to heavy-duty gas turbines is quieting the exhaust gas stream. The physical size of gas path silencing is directly related to wave length of the frequencies which are to be suppressed. Exhaust noise is believed to be created by flow of exhaust gases through the turbine, diffuser, transition, and flow separation over the bearing support struts; its spectrum is of a random broadband nature. Energy contained in the low frequency region, although not initially dominant, becomes dominant after the higher frequencies are attenuated. The silencers must, therefore, be large to accommodate long wave lengths associated with low frequencies. Costs involved with the manufacture and field erection match the physical size. CONVENTIONAL SILENCING Conventional techniques of reducing exhaust noise is by means of parallel baffles. Basically, the sound field is channeled through a duct of height, h, whose walls are lined with sound-absorbent material of thickness, t. In order to tune the silencer to the proper frequency range, h and t may be appropriately set. By increasing the length, L, greater noise reduction may be achieved. There are many approaches available by which these parameters may be calculated (1-3), 1 all of which overestimate the actual measured at- 1 Numbers in parentheses designate References at end of paper. A FLOW T EXHAUST FLOW ENTRANCE Fig. 1 Elements of conventional parallel baffles Fig. 2 Construction of Switchback 1

3 SHORT CIRCUIT SPEAKER Fig. 3 Original turning vanes Fig. 4 static acoustic test setup tenuation at operating conditions. However, once these physical dimensions have been determined, the number of channels necessary to provide an adequate flow area are arranged in parallel as shown in Fig. 1. Quite often, different baffle thickness may be required to extend the attenuation to include a broader range of frequencies. This necessitates arranging baffles and channels which are not always in alignment thereby degrading the flow field. It is false economy to increase the length of the silencer or number of sections without limit to achieve maximum silencing. Factors involved are increasing costs, back pressure, and regenerated noise. Greater back pressure reduces the power output of gas turbines (4) and, therefore, is not desirable. Regenerated noise is created within the silencer itself and is believed to be caused by flow interaction with the baffles. In some cases, regenerated noise is louder than the remaining exhaust noise at the silencer exit plane. SWITCHBACK SILENCER A major departure from this traditional silencing standard was initiated by the introduction of the Switchback silencer. A Switchback is a long duct of rectangular cross section folded back on itself one or more times such that 180-deg turns are formed along the duct axis. The walls of this duct are lined with sound-absorbent material similar to that used in the parallel baffles. Historically, Switchbacks have been used for silencing large jet engine test cells. A significant functional change was incorporated into this 2 basic design. This was a lined turning vane in the first 180-deg bend. A cutaway view showing the components and method of erection is shown in Fig. 2. Preliminary acoustic testing was performed on a full-scale model to finalize the design of the turning vane. This was accomplished by placing loudspeakers upstream and measuring the sound pressure level differential across the vanes. Originally, there were two turning vanes in parallel, Fig. 3, but preliminary acoustic testing told us that the acoustic energy was short circuiting through the narrow passage and thereby obviating the effect of the lined vane. When this passage was blocked off, the acoustic performance increased. The main design goal for the Switchback was for it to be a suitable replacement for the conventional parallel baffle silencers previously described. Total costs of the Switchback were not allowed to exceed those of conventional silencers. Since only three main pieces are to be erected, field assembly time and costs are substantially reduced. The acoustic performance, however, could not be compromised; i.e., the allowable sound pressure level (SPL) from the exhaust gas stream at 400 ft had to be equal to or less than 53 db (A). Another consideration which was factored into the Switchback design was its adaptability. Since two different models of heavy-duty gas turbines were to employ this new silencer, it was decided that GE MS 5001 gas turbines would be fitted with one Switchback, while the MS 7001 gas turbine, which has twice the weight flow of the MS 5001,

4 r mr 50 I c"e W Q ō _J C J 40 -J < 0 3e O- N OCTAVE BAND NUMBER OCTAVE BAND NUMBER Fig. 6 Levels at 400 ft calculated from estimated power at Switchback exit plane Fig. 5 Sound power level as a function position within the Switchback would employ two in parallel. By installing large dampers at the exit plane and in place of the wall marked * in Fig. 2, the unit could be used in conjunction with a heat recovery steam generator (HRSG) for a combined cycle-steam and gas turbine (STAG)-operation. This has the unique advantage of presenting a silenced supply of hot gas to an HRSG. STATIC ACOUSTIC TESTING In order to understand its acoustic behavior so as to be able to design a more efficient silencer, it was decided to test a fully constructed Switchback directly from the production line. Since conventional loudspeakers used in the preliminary testing were not capable of producing enough acoustical power to propagate through the entire silencer, a different test approach had to be attempted. The technique used was to mate a high intensity electro-pneumatic transducer directly to the silencer entrance by means of a horn. Since no such horns were commercially available, one had to be designed and constructed. The horn shape was chosen to be hyperbolic and designed to allow frequencies above 35 Hz to propagate. This would enable high amplitude low frequency acoustic signals to be generated, thus providing valuable information in the critical low frequency regime. The resultant test setup is shown in Fig. 4, Random broadband (pink) noise was introduced to the throat of the horn and the output at the mouth had a power of approximately 4000 acoustic watts (cf. conventional loudspeakers produce about 2 acoustic watts max). A microphone was scanned across planes perpendicular to the duct centerline at various positions along the flow path. These data were integrated on a real-time analyzer, thus yielding a true average power level at each station. The effect of the different duct sections on the sound power level generated by the horn can be seen in Fig. 5. Regions between curves show the attenuation of individual components of the Switchback. Attenuation indicated by region I is a result of a lined 90-deg bend and a 10-ft section of lined duct immediately downstream. A major portion of the attenuation especially at the lower frequencies takes place at the lined turning vane (region II). This is due to the large acoustically treated area which is exposed at normal or sharp angular incidence to the sound field. Due to the solid (reflective) walls of the second vertical section, no acoustic contribution was measurable for that 10-ft section which connects the two 180-deg turns. The second 180-deg bend (region III) performed almost as well as the one with the turning vane at the mid-frequencies, but not as well at the lower and higher frequencies. Finally, region IV shows the attenuation of 20 ft of a rectangular duct lined on three sides. By scaling the frequency of the spectrum 3

5 785 :(EN1) / 79 II 85 II II Fig. 7 SPL's measured 1 ft away from Switchback with respect to temperature and removing 5 db from the attenuation amplitude, a reasonable estimate of performance at operating conditions is determined. The 5-dB amplitude reduction is our best estimate for the effects of temperature on the lining's acoustical properties. Frequency scale factor is calculated simply by the square root of the absolute temperature ratio (5). This is to maintain the proper frequency wave length to duct dimension ratio. For example, the acoustic characteristics measured at, say, 500 Hz at 60 F would correspond to a frequency of 825 Hz at 1000 F. The resultant attenuation spectrum was then subtracted from the power of an unsilenced heavy-duty gas turbine and extrapolated to the standard distance of 400 ft. This expected spectrum is shown in Fig. 6. This method of analysis assumes that all the noise measured at 400 ft emanates from the exhaust plane. ACOUSTIC FIELD TESTING Final evaluation of the acoustic performance of the Switchback, however, could only be measured in the field. A thorough series of acoustic measurements were made on two different models of heavy-duty gas turbines fitted with Switchback ex- 0..,Cz 30 z GAS PATH SCALED r- GAS PATH MEASURED _ TRANSMISSION LOSS OF WALLS OCTAVE BAND NUMBER 4. Fig. 8 Attenuation of gas path and transmission loss of walls haust silencing. Data from these tests proved to be remarkably similar; their results are combined to form one analysis. It was found that acoustic measurements 1 ft out from the edge of the exit plane at both end and broadside locations as well as their respective EN1 position [45 deg up from the centerline of the exit plane a distance of times the minor cross dimension (6)] were essentially the same. If the acoustic levels originated in the gas stream at the exit, noise levels would decrease more rapidly with distance. This indicates that the EN1 and 1-ft levels were caused by flanking from a common source. Sound pressure levels measured 1 ft away from the silencer are indicated on an outline of the Switchback -- Fig. 7. A-weighted levels are used since they are good single number indicators of the levels present. Lower sound levels are evident at the end wall of the silencer. The reason is that this region is shielded from package noise. It is also apparent that the noise from the stack exit is negligible relative to the other sources. It is not clear precisely how much noise at the exit was present in the gas path and how much emanated from other sources. Assuming the levels of the flanking noise 5 ft below the lip exist at the exhaust lip, it appears that the true gas path noise is on the order of 73 to 74 db(a). Correcting for the flanking and comparing to unsilenced gas turbine exhaust, the estimated acoustic performance of the Switchback has been determined (Fig. 8). This performance can be con- I 4

6 sidered the lower bound since data at the exit was Switchback modular construction reduces ereccontaminated by the inclusion of some flanking tion time and costs. The Switchback silencer noise. Note the attenuation does not degrade satisfies the original goals of a suitable replacement for some applications of conventional silenc- severely with temperature and gas velocity. A reasonable suspect for the origin of the ing. flanking noise is the plenum and transition. The calculated transmission loss (TL) of the walls is REFERENCES also shown in Fig. 8. As can readily be seen, there is more sound reduction associated with the gas path than with the gas path enclosure. Considerable near-field data confirms that the cause of the flanking is due to the sound radiating through the walls. SUMMARY Switchback exhaust silencing affords a different method of reducing gas turbine exhaust noise. Gas path noise reduction is extremely high thus allowing a minimum of acoustic emission from the exhaust plane. Noise radiated through the walls can be easily reduced by simple walls, with no effect on turbine performance. Regenerated noise created within the silencer is small enough as to be unmeasurable. 1 Wells, R. J., and Tatge, R. B., "Prediction of Silencer Performance Using Transmission Line Theory," ASME Paper No. 71-GT-8. 2 Beranek, L. L., ed., "Wrappings, Enclosures, and Duct Linings," Noise and Vibration Control, McGraw-Hill, New York, Power Plant Acoustics, Manual No. TM- 8059, Headquarters, Department of the Army, Washington, D. C., Gas Turbine Sound and its Reduction, NEMA, Publication No. SM 33, National Electrical Manufacturers Association, New York, Cummins, J. R., "Acoustic Characteristics of a Gas Turbine Exhaust Model," ASME Winter Meeting, 1973 to be published. 6 Harris, C. M., Handbook of Noise Control, McGraw-Hill, New York, 1957, pp