In this work, a centrifugal pump was tested for its vibration signatures under different operational conditions.  The two abnormal operating conditions studied here are air bubbles and cavitation.  A transparent plastic cover was used in the experiments to observe the cavitation.  We found that the pump has higher vibration amplitude in the axial direction than in the radial direction.  From the experiments, we also determined that a significant amount of air bubbles will significantly increase vibration component associated with impeller vane pass frequency.

Cavitation might excite high frequency structural resonance. It might also reduce the impeller vane pass frequency vibration. Although cavitation is less likely to happen on a slow speed pump, it will develop very fast if it happens. 

Introduction

A pump is a mechanical device used to move liquids. Mechanical energy is transformed into hydraulic energy at the pump. Pumps can be classified into two categories: displacement pumps and centrifugal pumps. In this work, a centrifugal pump was tested and studied. 

The essential elements of a centrifugal pump are (1) the rotating element, consisting of the shaft and the impeller, and (2) the stationary element, consisting of the casing, stuffing boxes, and bearings.  Figure 1 illustrates the single stage bronze centrifugal pump used in this work. This pump has a single rotating metal impeller. Liquid enters at the center and is thrown outward radially by centrifugal force. The five impeller vanes can clearly be identified in Figure 1.



One of the important phenomenon in a pump is cavitation. Cavitation occurs when the pressure of the fluid drops below the vapor pressure for the temperature of the fluid.  When this pressure drop occurs, whether it is a system pressure drop or a localized pressure drop, voids or cavities (bubbles) will form in the liquid.  These bubbles implode or collapse when the fluid moves through the impeller to the high pressure side of the pump, causing the impeller to erode.  These implosions tear out tiny pieces of the metallic surface near the implosion.  This can be very damaging and eventually the impeller will fail. Figure 2 shows a schematic representation of the cavitation process.



There are three common causes of vapor formation in a liquid:

  1.  Flow separation of a viscous fluid from its guiding
       surface due to a surface discontinuity.
  2.  The addition of heat to the fluid, raising its vapor
       pressure (boiling point).
  3.  Reducing the pressure of the fluid to below its
       vapor pressure.

One important term in pump theory is net positive suction head (NPSH).  NPSH is a measure of the difference between the total suction head and the fluid vapor pressure.  The concept of NPSH is closely related with cavitation.  For a specific pump, there are both the required NPSH and available NPSH. The required NPSH is the factory suggested value which must be maintained to prevent the occurrence of cavitation.  The available NPSH is the real pressure difference between the suction head and the fluid vapor pressure.

Experimental Setup

The pump was installed on the machinery fault simulator (MFS) as shown in Figs. 3 - 5.  The suction and discharge sides of the pump are fitted with pressure gauges.  The pump discharge is directed through a manual modulating valve and then a flowmeter back into the head tank. Two single axis accelerometers were glued on the pump in the radial and axial directions respectively.  The vibration data was collected by using a SpectraQuest software/hardware system.

Experimental Procedure

The experiments are categorized into two groups.  In the first group, the original brass pump was tested.  In the second group, the original brass pump cover was replaced with a transparent plastic cover to observe the liquid motion inside the pump.

Brass Cover Pump Experiments

First, the pump was running around 3600 rpm to check the integrity of the system.  Through the transparent hose connected to the pump suction end, we noticed that a significant amount of air bubbles were sucked into the pump.  The pointer on the pump discharge pressure gauge was vibrating.  We found that the air bubbles were created by the returned water hitting the water inside the tank.  Before the air bubbles were exploding and disappearing, they were sucked into the pump.  The vibration data was collected for both of the cases, with and without air bubbles.

Next, the tank discharge valve was turned 45 degrees to restrict the flow rate into the pump.  This caused the pressure on the pump suction end to drop.  This might cause the water to cavitate as discussed earlier.  The water vapor pressure under room temperature is 0.935 inHg. The atmosphere pressure is about 29 inHg.  In order to prevent water vaporization, the pressure of the pump suction end has to be higher than NPSHrequired. We did not have the exact NPSHrequired data for this specific pump. Generally, the NPSHrequired is decreasing with flow rate or pump speed.  On the other hand, the NPSHavailable is increasing with flow rate and pump speed.  As a consequence, it can be argued that the possibility of cavitation is much smaller for low speed pumps than high speed pumps.

As the supply to the pump was restricted, the flow rate dropped.  Because of the lower flow rate and smaller impact force as the water returning to the tank, no significant amount of air bubbles appear. Vibration data were collected and used for later comparison.

The speed of pump was then decreased to around 2400 rpm. We found that under this speed, the air bubbles did not appear anymore.  The water flow is proportional to the pump speed, that is, the higher the speed, the greater the flow.  Therefore, the flow rate under 2400 rpm pump speed is lower than that of under 3600 rpm.  The smaller impact force caused by the slower flow rate is not large enough to create the air bubbles.  Vibration data for the normal operating status and cavitation status were collected.

Finally, the pump was running at around 1200 rpm.  As was the case for the 2400 rpm speed, no significant amount of air bubbles were created.  Vibration data for the normal operating status and cavitation status were collected.

Plastic Cover Pump Experiments

The purpose of installing the plastic cover is to visually observe the cavitation phenomenon.  With the brass cover, we have no definite answer as to whether there is cavitation or not.  We can just give the best estimation we can.  However, with the transparent cover, we can determine the cavitation formation with full certainty, and therefore, positively correlate the vibration signatures with the cavitation situation.  The procedures for experiments with the plastic cover are similar to those of brass cover.

In the data acquisition process, the frequency limit was set at 20 KHz.  Twenty seconds of data were collected for each case.

Experimental Observations and Results

Brass Cover Experiments – The acceleration spectra are presented in Fig. 6 for pump speed of 3588 rpm without air bubbles and cavitation.  Figures 6 (a) and (b) display the acceleration spectrum in the pump radial and axial directions respectively.  The fundamental 1X component and its harmonics can be identified.  The fifth harmonic, which corresponds to the impeller vane pass frequency (because there are five vanes on the impeller), has the highest amplitude.  Moreover, two impeller vane pass frequency harmonics also have high amplitude.  A comparison of the amplitude of Figures 6(a) and (b) indicates that the pump has higher vibration in the axial direction. 



The acceleration in the radial and axial directions is presented in Figure 7 for pump speed of 3590 rpm with a significant amount of air bubbles forming in the tank.  A careful inspection of Figure 6(a) and Figure 7(a) indicates that with formation of air bubbles, the vibration component associated with impeller vane pass frequency increases significantly.  The vibration amplitudes of 1X and its other harmonics components do not change much.  A comparison of Figure 6(b) and Figure 7(b) suggests a similar trend.  An examination of Figures 7(a) and (b) indicates a higher vibration level on the pump in the axial direction.



The suction head pressure was dropped below atmosphere pressure by approximately 20 inHg in the cavitation test for pump speed 3595 rpm. There is a great possibility that cavitation will appear under this condition.

The acceleration in the radial and axial directions is presented in Figure 8 for pump speed of 3595 rpm with cavitation forming in the pump.  Because there is not a significant amount of air bubbles forming during the cavitation test, a comparison of Figures 8 and 6 is appropriate.  A careful inspection of Figures 6(a) and 8(a) indicates that there is a frequency component around 1600 Hz emerging in the cavitation signal.  In Figure 6(a), the background noise has an almost constant level, which does not show in Figure 8(a). The 1X and its harmonics components have similar amplitude levels in Figure 6(a) and Figure 8(a).  A comparison of Figure 6(b) and Figure 8(b) has the same conclusions.



The pump speed was reduced to around 2400 rpm.  The suction head pressure was dropped below atmosphere pressure by approximately 15 inHg in the cavitation test.  Again, there is a possibility that cavitation will appear under this condition.

The acceleration in the radial direction is presented in Figure 9 for pump speed around 2400 rpm.  Figure 9(a) presents the data spectrum for pump speed 2355 rpm without cavitation.  Figure 9(b) presents the data spectrum for pump speed 2360 rpm with a possibility of cavitation.  Similar to the cavitation case with pump speed around 3600 rpm, there is a vibration component around 1700 Hz for the case with cavitation possibilities.



The pump speed was then reduced further to around 1200 rpm.  By turning the tank discharge valve to restrict the flow rate, the suction head pressure could be dropped below atmosphere pressure by approximately 5 inHg in the cavitation test.  The pressure drop could not be increased further because of the low pump speed.  It is not likely that cavitation will happen.

The acceleration in the radial direction is presented in Figure 10 for pump speed around 1200 rpm. Figure 10(a) presents the data spectrum for pump speed 1166 rpm without flow rate restriction.  Figure 10(b) presents the data spectrum for the same pump speed with a flow rate restriction.  As expected, there is no significant difference between Figures 10(a) and (b).

Brass Cover Experiments

Experiments were performed at different speeds at three tank discharge valve positions.  The pump suction head and discharge pressures were read for all operating conditions. The complete pressure data is shown in Table 1.



From Table 1, it can be found that the NPSHavailable and pump discharge pressure are all increasing with pump speed.  Another observation from Table 1 and the experiments is that the NPSH range from the appearance of cavitation to fully developed severe cavitation is also increasing with an increase in pump speed.  For example, for pump speed 3600 rpm, the NPSH for cavitation appearance is -13 inHg, while the NPSH for severe cavitation is -20 inHg.  It has a 7 inHg pressure difference.  For pump speed of 3000 rpm, the NPSH for cavitation appearance is -20 inHg, the NPSH for severe cavitation is -21 inHg. The pressure difference is only 1 inHg.  Moreover, for pump speed 2400 rpm, the cavitation appears at -18 inHg NPSH, and the cavitation develops into severe cavitation very quickly.  This observation indicates that, although it is less likely for a slow speed pump to have the problem of cavitation, if cavitation does start, it will develop quickly into a severe condition.

Figure 11 illustrates the vibration spectra in radial and axial directions, respectively, for pump speed 3619 rpm with the tank discharge valve fully open and without air bubbles. Figures 11(a) and (b) present the spectrum of pump radial and axial vibration with 20 KHz frequency limit respectively. Figures 11(c) and (d) display the same spectra in the 1 KHz frequency range.



Figure 12 illustrates the vibration spectra in radial and axial directions, respectively, for pump speed 3616 rpm with the appearance of cavitation.  Figures 12(a) and (b) present the spectrum of pump radial and axial vibration with 20 KHz frequency limit, respectively. Figures 12(c) and (d) display the same spectra in the 1 KHz frequency range.



Comparing the corresponding subfigures in Figure 12 and Figure 11, Figure 12(a) and Figure 11(a) have the largest difference.  In Figure 12(a), there are several peaks emerging around 6 KHz with the characteristics of structural resonance.

Figure 13 illustrates the vibration spectra in radial and axial directions respectively for pump speed 3617 rpm with severe cavitation.  Figures 13(a) and (b) present the spectrum of pump radial and axial vibration with 20 KHz frequency limit, respectively.  Figures 13(c) and (d) display the same spectra in the 1 KHz frequency range.



In Figure 13(a), the peaks emerging around 6 KHz with the characteristics of structural resonance are clearer.  An inspection of Figure 13(c) indicates that the amplitude of the vibration component with impeller vane pass frequency (the fifth harmonic of 1X) has decreased significantly.  However, this phenomenon does not appear for the pump axial vibration.  The vane pass frequency vibration is still strong, as illustrated in Figure 13(d).

Figure 14, on the following page, illustrates the vibration spectra in radial and axial directions respectively for pump speed 3007 rpm with the tank discharge valve fully open and without air bubble.



Figure 15, on the following page, illustrates the vibration spectra in radial and axial directions respectively for pump speed 3010 rpm with the appearance of cavitation.  Figures 15(a) and (b) present the spectrum of pump radial and axial vibration with 20 KHz frequency limit respectively. Figures 15(c) and (d) display the same spectra in the 1 KHz frequency range.



Figure 16 illustrates the vibration spectra in radial and axial directions respectively for pump speed 3010 rpm with severe cavitation.  Figures 16(a) and (b) present the spectrum of pump radial and axial vibration with 20 KHz frequency limit respectively. Figures 16(c) and (d) display the same spectra in the 1 KHz frequency range.

Summary

In this work, a single stage centrifugal pump was tested for its vibration signatures for different operational conditions.  Pump vibration was measured in the radial and axial directions by accelerometers.  The pump was running under three different speeds, 3600 rpm, 2400 rpm and 1200 rpm.  Air bubbles, caused by the impacting of returning water with the water inside the tank, were observed under pump speed of 3600 rpm.  Cavitation was created intentionally by closing the tank discharge valve somewhat to drop the NPSHavailable below
NPSHrequired.

Several observations can be made tentatively based on the experiments.

  1.  The centrifugal pump has higher vibration amplitude in the
       axial direction than in the radial direction.

  2.  A significant amount of air bubbles will greatly increase
       vibration component associated with impeller vane pass greatly.

  3.  Cavitation might excite high frequency structural resonances.

  4.  Cavitation might decrease impeller vane pass frequency vibration.

  5.  While cavitation is less likely to happen in slow speed pumps,
       it will develop very fast if it happens.

Dr. Lin Liu has published extensively on machinery health monitoring and prognostics/diagnostics.  He is also an expert in the finite element analysis of aerospace structures, crack propagation simulation and modeling in gearboxes, and helicopter structural analysis.  He has authored over twenty technotes at SpectraQuest and numerous research articles in peer reviewed journals and proceedings.  Lin obtained a Ph.D. degree in Aerospace Engineering from the University of Maryland at College Station, Maryland.

Dr. Suri Ganeriwala is founder/president of Spectra Quest, Inc.  He has over twenty-five years of industrial and academic experience in machinery vibration diagnostics and control, signal processing, and viscoelastic materials characterization.  Suri has developed a unique method of instruction using the Spectra Quest Machinery Fault Simulator (MFS), which is his brainchild from concept to completion.  He has authored over thirty papers and articles in journals, magazines, and books.  He obtained a Ph.D. in Mechanical Engineering from the University of Texas at Austin.  Suri can be reached at suri@spectraquest.com
or (804) 261-3300.