Story of the CW Pumps

CW Pumps are Circulating Water Pumps used in Thermal Power Plants. There were five vertical pumps in a row.

These pumps are critical for plant operation. When these pumps don’t function effectively, the power output might have to be de-rated. For instance, a 500 MW plant might have to be de-rated to 180 MW.

In this case the CW Pump 1 was vibrating heavily, which was as high as 100 microns.

In order to analyse the problem the following data were examined:

1. Vibration data and historical trends for the machines

2. Signature analysis of the selected machines

3. Detailed inspection of the related systems at site

4. Acceleration time waveforms

5. Natural frequency test


1. Pump 1 experiences higher vibration when compared to the vibrations of other CW Pumps.

2. 2 x RPM harmonic dominates the vibration signature.

3. Vibration “water fall diagram” also indicates that the 2 x RPM harmonic varies.

4. Acceleration waveform is unsteady, i.e. never settling down to a steady state condition.

5. There are a number of natural frequencies of the pump, which indicate multiple degrees of freedom of the system.

6. One of the natural frequencies is very close to the two times operating speed of the pump.

7. at about 60 meters upstream from the pumps the channel took a sharp bend. This obviously creates a variable velocity and pressure gradient across the inlet channel. The effect of this on the pumps is very interesting indeed. It means that while Pump 1 receives the water at a higher velocity than Pump 3, Pump 3 receives the water flow at higher pressure than Pump 1. The effect of this was very vivid. As water leaked out of the gland packing of Pump 3 due to higher pressure, Pump 1 behaved more gently. In addition, if I consider the force generated due to angular momentum of the impeller it makes the situation worse. Since the impeller rotates in the anti-clockwise direction the force is directed upwards towards the gland packing (remember that the pump is a vertical pump). Add to this the Coriolis effect, which makes the situation even worse.Velocity and pressure gradients across the flow was observed in the intake channel and in the bay area. The pressure at the Pump 1 inlet is the highest among all the pumps, while it inlet velocity is the lowest among all the pumps of the set. This is the cause of the abnormal vibration and behavior of Pump 1.

8. Now I turned my attention to the water inlet channel. This channel was running in the East-West direction, which meant that the water flow would be inherently aperiodic or non-harmonious (Coriolis effect) which takes effect when water strikes the vanes of the impeller. Unfortunately, perhaps due to oversight, aperiodicity was inadvertently built into the design. This non harmonious behavior could be seen up to the condenser, exhibited by the asynchronous fluctuation of the inlet and outlet pressure gauge readings.

Discussion and Analysis

1. Clearly this is also a non-linear vibration problem.

2. The cause of the problem is high pressure of water flow at the pump inlet.

3. Hence the solutions lie in reducing the pressure at the inlet of the pump

4. No mechanical problems of the pump were observed.


1. Placing of a diverter in the flow path approximately at a place right after the bend of the channel at about 45 degree in direction of the flow. The idea is to obstruct and divert the flow towards pump 1. This may be done by placing a barrier which would be around 3 meters in length, 0.5 meter in width and 1 meter in depth. The details have been discussed at site.

2. Lifting the pump by 2/3 mm so as to decrease pressure and increase flow. This has been discussed at site.

3. Placing an aerodynamically shaped splitter in front of the pump inlet as provided for CW Pumps of Stage I. This would increase the inlet velocity and decrease the pressure at the inlet of Pump 1. This has also been discussed at site.

In this case option number 2 was implemented.


Vibration Signatures:

Before rectification: on 29th July 2016

Vibration Spectrum as on 29th July 2016, 1x = 223 microns

After rectification: on 28th August 2016

Vibration Spectrum as on 28th August 2016 1x = 9 microns.

Comparing the two spectrums we see that the drop in vibration level (comparing 1x fundamental frequency) was of the order of 223/9 = 25 times (approximately).

copyright © Dibyendu De