• The noise experienced is very similar, if not identical, to classical cavitation (sounds like pumping marbles).
   
•  The phenomenon can occur with either a forced draft or induced draft cooling tower.
   
•  The resulting noise tends to be more prevalent on negative suction pressure systems but will occur on positive suction pressure as well.
   
•  The introduction of small amounts of air to the pump suction often reduces or eliminates the cavitation noise.
   
•  Entrained air introduced as Number 4, above, has little effect on the pump's life expectancy.
   
•  Such small amounts of entrained air have little deleterious effect on other tower system components; however, each system must be analyzed for possible harmful effects.
   
•  Unlike classic cavitation, throttling of the pump discharge to a lower capacity usually has little impact on the noise level.

In an effort to determine the probable cause of the noise, we made several visits to sites experiencing such problems. Typically a detailed site inspection is conducted and an audible recording made of the noise spectra for laboratory analysis. Subsequent analysis revealed the following:

• No distinct frequencies were found.
• The predominant noise measured was broad band, occurring above 300Hz.

There are two mechanisms for generating pump noise: liquid and mechanical. Both sources produce acoustic pressure fluctuations that can be transmitted as audible noise.

For centrifugal pumps, mechanical noise is generally the result of component imbalance (impeller and/or coupler), coupler misalignment, rubbing components or improper installation of the base plate and/or motor. These mechanical mechanisms generate distinct frequencies equal to rotational speed and/or multiples (1,2,3) of rotational speed. Because the noise spectra did not reveal distinct frequencies, we concluded the noise was not mechanically generated.

The second mechanism for generating noise is velocity of the liquid entering the pump. Liquid noise is directly produced by water movement and is fluid dynamic in character. Turbulence, flow separation (vortex), cavitation, water hammer, flashing and impeller interaction with the volute cutwater are all examples of fluid dynamic noise sources.

According to the Pump Handbook, 2nd Edition authored by Igor J. Karassik, there are generally four types of pulsation sources in pumps that result from liquid noise:

• Discrete frequency components generated by the pump impeller
   
•  Broad-band turbulent energy resulting from high flow velocities
   
•  Impact noise consisting of intermittent bursts of broad-band noise caused by cavitation, flashing and water hammer
   
•  Flow-induced pulsations caused by periodic vortex formation when flow is past obstructions and side branches in the piping system

Discrete frequencies, item I from above, can be ruled out as the cause of the noise problem. As previously mentioned, distinct frequencies, such as the vane passage frequency and/or its multiples, were not found. This would be the case if an interaction occurred between the impeller and volute cutwater.

Item 2, 3 and 4 are generally identified as broad-band noise and would occur in the 300 Hz and above frequency range as identified on our noise spectra. Therefore, we believe the noise was being generated by one or more of these liquid sources.

The pump noise observed is that of cavitation. Pump cavitation results from the formation of vapor bubbles when the localized static pressure is lower than the vapor pressure of the liquid being pumped. To evaluate the pump for classic cavitation (NPSHR greater than NPSHA) close the discharge valve thus pushing the pump back on its curve toward shut-off. The noise should diminish significantly if it originated from classic cavitation, as lower pump flows require reduced NPSHR. If the noise continues, entrained air may be the cause.

We knew classic cavitation was not occurring in one particular installation because the operating suction pressure measured 30 feet above vapor pressure. Thus, the NPSH available was approximately twice that required by the pump. For that reason, we knew the pump is not cavitating because of insufficient NPSHA. SO, if not classic cavitation,what was causing the noise?

It is a well-documented fact that highly aerated cooling tower water can contain as much as 4-6% excess air. This excess air exacerbates the potential for a noisy pumping installation. The excess air absorbed in the cooling tower comes out of solution as it flows through the piping and becomes entrained air. Suction velocities often are high enough to pass the air through. However, air sometimes collects in an area of the suction piping and creates an obstruction (Figure 1). As the liquid passes through this reduced area, its velocity increases, creating an area of reduced pressure. At this point of reduced localized pressure, water vaporization occurs with the resulting bubbles passing into the pump impeller where they collapse, and produce "cavitation."

Several noise control techniques have been successfully employed in the past to reduce excessive noise:

• Increase or decrease the pump speed to avoid system resonances of the mechanical or liquid systems.
   
•  Increase liquid pressures (NPSHA, etc.) to avoid cavitation or flashing; decrease suction lift. This could include raising the tower, lowering the pump, or straightening the suction piping (see below) to reduce friction losses.
   
•  Modify the pump so that the clearance between the impeller diameter and casing cutwater (tongue) or diffuser vanes is increased.
   
•  Inject a small quantity of air into the suction of a centrifugal pump to reduce cavitation noises by providing a shock absorbing cushion to minimize the impact of recondensation of water vapor within the pump impeller.

The injection of small amounts of air can usually be accomplished quickly and easily in the field with minimal expense. Small amounts of entrained air usually cause no problem in the cooling tower/condenser circuit. B&G therefore considers this alternative desirable and recommends its application as a solution to certain field problems or, at a minimum, as an analytical tool.

Contributing Factors In addition to the techniques outlined above to reduce or eliminate noise, attention must also be given to two other factors that can exacerbate noise problems: vortexing of liquid in the tower pan, which is the most common source of entrained air, and the suction piping arrangement itself.

Vortexing of Liquid in the Tower Pan The amount of entrained air caused by vortexing depends on several variables, but particularly the vortex size and the submergence level of the pump suction pipe below the water level of the pan. The most common method of eliminating vortexing in the tower pan is by the inclusion of baffle assemblies to eliminate the formation of vortexes. Raising the fluid level in the pan to sufficient depth also can resolve this problem.

Suction Piping Coupled with the vortexing phenomenon, or by itself, the improper layout of the pump suction piping can be a significant contributor to the generation of pump noise.

Friction losses caused by undersized suction piping can cause an increase in fluid velocities into the pump. The pipe friction can be reduced by using pipe sizes that are one to two sizes larger than the pump suction nozzle. Suction line velocities should never exceed 10 ft./sec. and should realistically be designed for 5 ft./sec. Suction headers should be designed for a maximum velocity of 3 ft./sec.

Concentric reducers used to step down to the pump flange from the larger suction piping can also be a culprit if improperly installed. At one problem facility discussed earlier in this article, the reducer was installed upside down (Figure 1) with the flat side on the bottom. If the pump liquid contains air (or vapor), as it did in that case, the air can become trapped in the sloped area of the reducer now located on "top." As a minimum this will obstruct the flow passage causing higher velocities and thus localized vaporization. If transported into the impeller, the trapped air can create a momentary choking that could even cause shaft breakage.

Elbows used on the pump suction flange, while convenient, can cause an uneven flow of liquid into the impeller when the elbow bend is along the axis of the pump shaft. If the elbow is a short radius design, you may create enough turbulence to produce entrainment which can, and does, exacerbate noise problems. The addition of a second elbow only increases the problem, especially if it has been added in a position at right angles to the first. Numerous technical publications, as well as the Hydraulic Institute itself, state that a minimum of five pipe diameters of straight run of pipe should be provided before the pump suction flange to allow for a smooth, unimpeded flow to the impeller.

System strainers need to be located on the discharge side of the tower pumps, and not on the suction side. On a different project, the location of basket strainers directly in front of the suction flange on a large HSC pump resulted in an unexpectedly high pressure drop. This turned out to be one of the contributing factors to poor pump performance in that installation, as well as higher noise levels from the pump.

It must be understood that each job site has its own particular set of operational requirements and, therefore, there is no single solution to the noise problems. The Hydraulic Institute is considering an HI standard design recommendation to allow an ample margin of safety between NPSH available and the pump manufacturer's published NPSH requirements. The margin of safety would be a minimum of 1.7 times NPSH required or NPSH required plus 5 feet, whichever is higher.

If you are experiencing noise problems and need some help diagnosing the cause, contact your local Bell & Gossett Representative.