Murray 2 power station was built by the Snowy Mountains Hydro-Electric Authority (SMA) in 1966-68. Its 269m head feeds four 137MW vertical shaft Francis turbines rotating at 333.3rev/min. The hydraulic conduit system consists of a lined concrete tunnel 2460m long (7.47m horseshoe section), an in-ground surge tank and two surface pipelines — 1550m long and 4.2m in diameter. Each pipeline bifurcates into 2.5m diameter pipes, with 2.5m diameter spherical valves — the main inlet valves (MIVs). As the machines are numbered 11, 12, 13 and 14, the pipelines are numbered 11/12 and 13/14.
Since commissioning, the pipelines have emitted a low frequency noise whenever any of the units was generating. Following complaints from a local resident, the noise was investigated.
The turbine runner has 17 blades and a blade pass frequency of 94.4Hz. Spectral analysis of the noise showed a 95Hz spike, with occasional 94.7Hz lower power spikes (the system was beating). The blade pass frequency was exciting a natural frequency somewhere in the water passage, causing resonance in the pipe system. The resonance was caused by the minor flow blockage as a turbine runner vane passed a fixed point, causing a minute pressure wave. The length from the first stay vane to the MIV was about the correct length to allow a wave travelling upstream to reflect from the MIV back to the runner, arriving just as the next runner vane passed and reinforcing the next pressure wave. The MIV is a spherical valve whose bore equals the pipe diameter, and it has no surfaces to reflect pressure waves, but the massive cast steel structure acts as an impedance change and therefore a reflector. In effect, it meant the pipelines acted as loudspeakers.
The pipelines traverse a hill slope about 300m uphill of a local resident who complained of the noise. From the rear of his house, the pipelines are visible over an arc of about 120°.
Noise levels along the pipeline were measured at 94-102dBLin at 100Hz with a 1/3 octave filter. Analysis, filtering outside the range 90-100Hz, showed a virtually pure 95Hz spike with no significant noise at any other frequency in that bandwidth.
The noise level surged typically over 3-4min due to the minor mismatch between the exciting frequency and the natural frequency.
The noise was offensive when standing beside the pipelines; after several hours, it caused discomfort. Apart from a very slight 95Hz vibration, no movement of the pipelines was detected. A velocity transducer on the pipeline surface recorded 0.1m/s peak velocity. The water pressure fluctuated by ±5kPa RMS at the noisiest part of the pipeline. The resonance is not likely to cause fatigue of the pipelines.
Environmental legislation limits noise in this region to 5dBA over ambient. Background noise at the local’s house averaged 60dBLin (about 45dBA) by day. At night, it fell to below 50dBLin (about 35dBA). When the units were on line up to 75dBLin (about 60dBA) was recorded, the majority of the extra noise being pure 95Hz monotone. During the day, the noise could barely be heard, but at night when the background noise levels were low, it was clearly audible and offensive.
The noise was nodal: moving along the pipeline over 5m the noise would decrease from 95dBlin to 75dBlin and rise to 95dBlin. These nodes were caused by standing waves generated in the pipeline and the location remained essentially constant regardless of unit or pipeline loading. Even when only one unit was on line, the other pipeline would register the same noise levels. The disturbance was travelling up the line, reflecting from the surge tank and travelling back down both pipelines. In fact, the noise levels were slightly lower if two units were operating on one pipeline instead of just one unit. The second unit seemed slightly to interrupt the resonance.
There were five high nodes along the length of the resident’s house, the worst being directly outside his bedroom. A noise survey at the house showed that when any units were on line, the noise increased by 10-15dBAlin, almost totally caused by the 95Hz noise. It was offensive to the local and warranted further action.
Examining the options
Several ways to eliminate the noise were excluded:
•Erecting noise barriers along the pipeline was not feasible because the 120° field of view of the pipelines at the house and the nature of the noise precluded any effective barrier.
•Typical noise barriers at the resident’s house, eg trees, would not be effective because of the noise’s low frequency and its continuous drone.
•The pipelines cannot be buried because they are not designed for external pressure and because they have expansion joints which require regular maintenance.
•To enclose the pipelines, a massive structure would be required. Suppressing the noise outside the power station is possible: an in-line silencer or a side branch resonator could be fitted to each pipeline. However it would be very expensive and installation would involve a total station outage lasting over at least four months.
Investigation focused on suppressing the noise at source. Since the noise source was pure 95Hz pressure fluctuations the best approach was to provide equal but opposite pressure fluctuations to cancel out the original disturbance. Passive and active means were investigated.
For passive suppression a ‘quarter wave side branch’ pipe allows some pressure waves to be reflected in the branch as a positive wave, returning to the main pipe as the negative portion of the wave passed to cancel it out. This was not feasible at Murray 2.
Alternatively, a Helmholtz resonator is a water-filled device which resonates and produces pressure pulses into the main pipe which are 180° out of phase. A trial resonator was made using a 20L vessel bolted onto an existing 80mm NB valve attached to the spiral casing and filled with water. The resonator’s frequency was adjusted by varying the ratio of the mass of water in the chamber to the mass of water in its nozzle. A tuning sleeve inside the nozzle pipe was screwed up or down to alter the effective length of the nozzle section and hence the ratio of masses.
When the frequency was 95Hz, it resonated strongly and produced pressures inside the chamber up to ±1000kPa. The initiating disturbance in the pipeline was only ±10kPa at the tapping point. The effect on the pipeline was a dramatic reduction in the noise levels, but there was insufficient energy in the resonator to overcome the beating phenomenon and after a few minutes the pipeline noise resumed its previous intensity. The resonator gave sufficient confidence to pursue the design further.
The solution
Following the success of the small unit, a larger resonator (270L), with similar chamber, nozzle and tuning sleeve, was fitted to the spiral casing manhole of unit 14 turbine. For the best results, the resonator should be installed at the point where the pressure disturbance is at its maximum. This was not possible because the spiral casing was made from heat treated steel plate, and the preferred location was embedded in concrete. It was simpler to choose the spiral casing manhole, even though it was only 1m from the probable reflecting surface. The manhole had a large diameter, allowing considerably more energy to be imparted to the flow than with the 80mm trial unit.
The resonator was installed on unit 14 manhole and tested. Pressures inside the chamber were measured, and gave amplitudes of only ±50kPa. Although it was not resonating at the correct frequency, it was already suppressing the disturbance in the pipeline. As the initiating disturbance was lessened, the resonator pressure fluctuations had been lessened.
There was a marked reduction in noise levels along the pipeline. With no resonator and only unit 13 on line, loaded to 100MW, levels of 96-100dBLin were recorded beside the pipeline. With the resonator fitted, the load on unit 13 transferred to unit 14 and unit 13 MIV closed, the noise level dropped by 25dBLin. Greater reductions — up to 35dB — were observed but not sustained.
In practice, there was a much greater reduction in the 95Hz noise. Measurements were made with a Larson-Davis 800 noise level meter set to 1/3 octave filter with the mid point set at 100Hz. Since the meter was averaging the noise level over a band of 88.5-111.5Hz and most of the energy was a 95Hz spike, the observed 25-35dBlin was probably more like 35-45dBLin. With coarse frequency steps the exact drop in noise levels could not be measured and must be estimated. However, the harsh nature of the noise was greatly reduced, and it became more of a white noise than a monotone drone.
The 95Hz component of the noise was barely audible at the local’s house when the resonator was brought into service. When unit 14 was on line with another unit, the resonator did not operate as well as when unit 14 was on line by itself, and when unit 14 was off line and its MIV closed, the resonator had no effect.
Standing beside the pipelines, the 95Hz noise can still faintly be heard. The noise levels are about 65dBLin at 100Hz, or 52dBA. Background noise levels at the time of measurement were about 20dB lower than these levels.
Noise levels inside the power station near the turbine were reduced by 2-3dBLin at 100Hz and the 95Hz noise could not be clearly heard inside the station above the turbine noise. Overall noise levels are still relatively high and further work is being undertaken to reduce them.
The 95Hz noise is still faintly audible at the local’s house. This is probably because we have a 450mm ‘cannon’ firing pressure waves at right angles into a 2500mm diameter pipeline. The limited diffusion angle makes it impossible to treat the whole cross section of the pipeline. However, the results are a great improvement. By day, the noise levels are below background levels, but at night, the background noise falls to about the same level as the pipe noise at the house and the pipe can still be faintly heard.
Three resonators were ordered for the other turbines and noise levels are now considerably lower. As the resonators replace the manhole covers in the spiral casings no cutting or welding was required and access for regular maintenance remains unaffected.
Since the reson-ators could generate pressures theoretically up to twice the maximum static pressures, their design had to use twice maximum static pressure as the design pressure. Further, since they resonate, fatigue is a major factor to consider and they were designed for an infinite number of pressure cycles.
Practice has shown the vessels to be overdesigned, but for the safety of personnel and the station, these criteria had to be followed.