THE PERILS OF INSTALLING VFD DRIVES ON VERTICAL PUMPS

Nelson L. BaxterABM Technical Services

Mooresville, In.NelsonBaxter@Att.net

Abstract: In order to save energy and to have better flow control, variable frequency drives are often installed on equipment. For the above reasons, vertical pumps are having VFDs installed at an increasing rate. The application of VFDs on these vertical systems can result in motor insulation and bearing problems and high levels of structural vibration. This paper discusses issues to consider when installing a variable frequency drive with the main emphasis on structural vibration problems on vertical pump VFD applications.

Key Words: Natural frequencies; dynamic stiffness; amplification factor; VFDs; asymmetric stiffness; mode shapes

Introduction: It is well known that the installation of VFD drives has the possibility of introducing vibration problems on rotating equipment. When a machine runs through a wide speed range the probability of operating at or near a resonance is significantly increased as compared to a machine operating at a single frequency.

VFDs are generally utilized to slow the speed of a machine down to match the load, thus saving energy and allowing better control of the process. The inherent nature of the design of vertical pumps means that there is a large mass (the motor) sitting on top of a relatively narrow structure (the vertical pump case). That combination of a heavy mass on a relatively weak support results in a low natural frequency. Most vertical pumps therefore have a natural frequency that lies near or below their operating speed. Therein lies the problem. When the VFD lowers or raises the speed of the vertical pump to match the load, there will most likely be a speed encountered where the rpm matches a structural natural frequency. Since the structures are generally fabricated steel, the damping is low and the amplification factor is therefore high. A residual unbalance force that might have resulted in a low amplitude of vibration for a non-resonant system can therefore produce very high levels of vibration when a resonant speed is encountered on a lowly damped vertical pump. To complicate things even further, vertical pumps usually have non-symmetric stiffness values in the two orthogonal directions that correspond to the directions that are in line with the discharge line and 90 degrees out from the discharge line. This means that there can often be two natural frequencies to deal with instead of one. Threading the needle so that the desired operating range of the pump does not excite either of these natural frequencies can pose a challenging problem.

There are also electrical harmonic issues and bearing fluting problems that can arise when VFDs are installed. This paper discusses test techniques to identify vertical pump natural frequencies and possible solutions to this class of problem. Included in this paper are also short discussions related to the aforementioned electrical issues associated with VFD applications.

Background on Variable Frequency drives: Variable frequency drives were developed as alternatives the DC Drives or variable speed hydraulic couplings as a means of controlling the speeds of rotating equipment. Variable frequency drives use standard AC induction motors, with modified winding insulation, and obtain the variation in speed by varying the frequency of the current supplied to these motors. The first step in the process is to convert the incoming AC signal to DC. By using rapid switching devices, this DC current is converted back into variable frequency AC current. This rapid switching process introduces a variety or problems including bearing fluting type failures, stressing of insulation and overheating of electrical components. Figure 1 shows the components of a Pulse Width Modulated variable frequency drive. To the left is the incoming 3-phase AC source. The middle section is the rectifier. This section converts the signal to DC current. The right section is the inverter section. It chops the DC signal up and reconstructs it into an AC signal, the frequency of which can be varied. The resulting synchronous no load speed is:

RPM=120 XFrequencyNo . Poles (Equation 1)

Plot from Electrician’s Technical Reference-Variable Frequency Drives by Robert S. Carrow

Figure 2 shows the output from an inverter. Note that the DC signal is rapidly turned on, then back off. There is a variation (modulation) in the amount of time the signal is left on (pulse width). This type of frequency converter is therefore referred to as a Pulse Width Modulation (PWM) inverter. The variation in the conducting time periods is what produces the resulting current wave form shown in the bottom plot.

Narrow pulses produce low current flows.

Wide pulses produce high current flows.

Figure 1

Figure 2

Note the saw tooth pattern of the resulting sine wave. As can be seen, the inverter does not produce a pure sine wave. However, by the use of faster switching speeds and filters, the signals can be made to more closely approximate a sinusoid. The closer the current can be made to look like a pure sine wave, the quieter and more efficient the motor will run. Better efficiency also results in less heat being produced. Herein lies a contradictory problem. Faster switching speeds increase efficiency and reduce noise. However, the down side to faster switching speeds is that the faster the change in voltage, the higher the level of voltage spikes that are produced. These voltage spikes, which can be much higher than what a motor would see on a normal across the line application stress the insulation. A small nick in the insulation or other damage that might not result in any detrimental effect in a standard application can result in failure of the insulation on a VFD application due to the aforementioned voltage spikes introduced by the rapid switching of the inverter. If we go back to basic electronics, the voltage produced in an inductor is

V=L dΦdt . Note that the voltage level is a function of the inductance, which in this instance is a

function of the length of the cable between the drive and the motor, and the rate of change of the flux. Two key factors in determining the likelihood of problems are therefore the length of the line and the switching rate of the inverter. This brings up an interesting scenario where an inverter driving a motor may work fine at one location, then have problems at another site because the length of the line between the inverter and motor is longer at the second location.

The above information was taken from an earlier paper by N. Baxter on variable speed drives,From a strictly electrical viewpoint, it can be seen that conversion to a VFD drive, particularly on a motor that was not originally designed for VFD service, can result in premature failure of the motor windings as well as bearing fluting problems.

Mechanical Considerations: There are three primary mechanical factors that need to be considered when converting a vertical pump to VFD service. The first involves matching the flow conditions. This is particularly true if other pumps that remain operating at a fixed speed are left feeding the same header. If for instance, there are two pumps operating at full speed and then a third pump speed is backed down with a variable speed drive to fine tune the flow, it is possible that the third pump might not create enough pressure to produce any flow and would therefore become dead headed below a certain speed. System back pressure versus flow must be compared with the flow head curve of the pump that is going to be put into VFD service. Because of this flow issue, there will be limits as to how low a speed the VFD driven pump can operate. A second consideration involves acoustical natural frequencies. When a pump operates

Note that the amplitude of the reflected signal is much higher than the amplitude of the original square wave.

Figure 3

at a constant speed, the vane pass frequency is a fixed number. Lengths of pipe can be determined that will insure that there are no acoustical natural frequencies in close proximity to the vane pass frequency. When a pump is put on a VFD, then the number of vanes (usually two to seven) times the rpm can sweep through a large frequency range, so the chance of exciting an acoustical natural frequency goes up significantly. While the above electrical issues, acoustical natural frequency problems and even torsional problems can occur, by far the most common problem encountered when converting a vertical pump to a variable speed application is the potential to operate at or in close proximity to a structural natural frequency.

Determination of Vertical Pump Natural Frequencies: There are two commonly used procedures for determining vertical pump natural frequencies. The first is a startup or coast down test. In order to perform this test, a once per revolution tach signal is installed on the pump shaft. The pump is then started and the amplitude and phase are monitored using tracking software during the startup and coast down process. In reality, when a pump is started across the line it is difficult to utilize this test technique. The problem that occurs is that the pump will get to speed so quickly that there is not enough time to obtain an adequate data set. Pumps, unlike fans or turbines, because of the drag of the liquid being pumped, also tend slow down very quickly, so coast down tracking tests are also seldom successful. If a pump is already on a VFD, and the rate of speed change can be controlled, then the startup & coast down tests results will result in reliable data.

The most common method of determining structural natural frequencies on a vertical pump is to perform an impact test. This test requires that the pump be impacted and then a spectrum of the response be produced as the vibration signal from the impact decays. The impact can be produced by a block of wood or a large hammer that has a soft tip. This simple test will provide a response curve that shows the natural frequency of the pump. Resonance tests need to be performed both in line with the discharge pipe and in the direction that is 90 degrees from the discharge line (See Figure 4). While this simple test provides some useful information, a more comprehensive test can be performed using an instrumented hammer to excite the pump structure. When an instrumented hammer that provides a force signal is utilized in combination with a multi-channel spectrum analyzer, it is possible to determine the dynamic stiffness of the structure(Figure 6), the mode shape(Figure 5) and also measure the coherence between the excitation force and the response. Figures 4 & 5 show a typical vertical pump and its first bending mode.

Figure 4- Measurement Directions Figure 5-First Bending Mode In-Line 750 Cpm

Dynamic Versus Static Stiffness: Figure 4 shows a typical vertical pump. The pump is statically much stiffer in the direction that is in line with the dischage pipe. The discharge pipe tends to stiffen the pump’s first bending mode in the direction direction that is in line with the pipe. Located tn the direction that is 90 degrees out from the dicharge line is usually an opening that is cut into the pump case. This opening allows access to the coupling and the pump seal. That opening lowers the static stiffness in the 90 degree out direction. Because of the stiffening effect of the discharge pipe and the weakening effect of the access opening, the natural frequencies are different in these two directions. Figure 6 shows the stiffness as a funciton of frequency taken from a vertical pump.

In Line with discharge pipe

90 degrees out

Figure 4 Figure 5

Figure 6

Cutout opening to

get to coupling and seal

Discharge line

There are two regions of interest in the data shown in Figure 6. The first is the static stiffness which is projected out to o Hz. The static stiffness values for this pump were the following: The static K value in line with the discharge pipe at the top of the motor was 31,000 lb/in. The static K value 90 degrees out from the discharge line at the top of the motor was 22,000 lb/in. The static stiffness values are in agreement with the observation that in line with the discharge line, the discharge pipe stiffens the system and 90 degrees out the cut out for access to the coupling weakens the structure in that direction. The next stiffness of interest is at the natural frequency which is very close to the operating speed. The dynamic stiffness value in the in line direction at 750 cpm was only 8,000 lb/in. The dynamic stiffness value in the 90 degree out direction at 750 cpm was 25,000 lb/in. As can be seen, in the direction that had the greatest static stiffness, the dynamic stiffness at the natural frequency was less than a third as much as it was in the direction with the lower static stiffness. This explains why the vibration on a vertical pump can be much higher in the direction that would appear to be the stiffest.

Case History : A water company had plans to install a Variable Frequency Drive on one of its vertical pumps. While on site doing analytical work on another pump, impact tests were performed on the pump that was to have the VFD installed. Figure 7 shows that the pump, which ran at 900 rpm, had a natural frequency at 1125 cpm in the direction that was in line with its discharge pipe and a natural frequency of 750 cpm 90 degrees out from the discharge line.

Since the operating speed was located between the two natural frequencies, there had never been a vibration problem with this constant speed pump. The issue was that the water company was planning to put a VFD on this pump. The frequency response data in Figure 7 showed that if the speed were lowered by a VFD that the vibration would significantly increase as it approached the 750 cpm natural frequency. This information was conveyed to the water company, but they said it was too late to do anything because the VFD was already ordered.

1125 CPM In line with discharge

pipe750 CPM 90 degrees out from discharge

pipe

Pump Speed

900 RPM

Figure 7 Figure 8

After the installation of the VFD, the pump was tested to determine its response as the speed was lowered. Figure 9 shows the results of the response as the speed was moved down.

As predicted by the natural frequency test (Figure 7), the amplitude increased significantly as the speed was reduced by the VFD. Based upon both the natural frequency test results from the impact test and from the startup-coast down data, it was determined that modifications were necessary if this pump was going to operate on a VFD.

Solution:The first step in determining a solution was to calculate the system damping. The

following equation which calculates the amplification factor at resonance using the rate of phase shift was used to calculate the damping.

Q= π NC∆θ

360 X ∆ F (Eq. 2 ) For the phase data in Figure 9, the Q value was 23.5

ξ= 12Q (Eq. 3) This results in a damping ratio = .021

Two assumptions were then made. The first was that the addition of mass to the motor would not change the damping. This is a reasonable assumption because nothing in the supporting pump column was being modified. The second was that the motor acted as a lump mass. This was also felt to be a reasonable assumption because based upon several other vertical pump mode shape projects, the motor on a vertical pump generally acts like a single solid mass.

The next step in coming up with the amount of mass that would be required was to determine how much lower the natural frequency would need to be shifted in order to allow the pump to run below the Hydraulic Institute Standard acceptance level, which in this case was 6.0 mils. Using the response ratio equation shown in Equation 4 and an original value of 15 mils, the calculations indicated that if the natural frequency could be shifted down 10%, then the pump at the lowest speed required by the process would run below the Hydraulic Institute level. This meant that the motor mass needed to be increased by 21%.

X /X0=1/√¿¿ (Eq. 4)

Figure 9Phase

Amplitude14 mils 785

CPM

Figures 10 and 11 show the predicted response ratios without (Figure 10) and with (Figure 11) the 10% additional mass. Both curves use the same damping ratio which as mentioned previously was determined by the startup-coast down curve’s rate of phase shift.

CALCULATED RESPONSE RATIOS WITH NO MASS ADDED

CALCULATED RESPONSE RATIOS AFTER ADDING 21% MORE MASS

785 Rpm

750 Rpm

900 RPM

750 RPM Expected Response 6 mils

900 RPM

With the as found natural frequency being at 785 cpm, the VFD would have to be set to not run between 750 and 825 RPM. That would mean that the pump could only be set to drop in speed from 900 down to 825 RPM.

With the addition of the 21% extra mass, calculations show that the natural frequency in the 90 degree out direction should drop to 715 RPM. Per the calculations, the pump should be able to operate down to 750 RPM before it hits the Hyd. Inst. alarm level of 6 mils. Note- This assumes that the amount of unbalance does not increase.

Expected operating range passes through peak of resonance curve.

Expected speed range is above peak in resonance.

Figure 10

Figure 11

Special mounting brackets were designed and installed on the pump. Segmented plates cut out with an industrial laser that were small enough to be handled by two men were then bolted to the brackets. A total of 3100 lbs. was added to the motor. Figure 13 shows the motor with the weight installed. Following the installation of the additional mass, an impact test was performed to evaluate the change in the natural frequencies. Figure 12 shows the response curve. Actual Response Curve Following 21% Mass Increase

Final Results: Before Mass After MassIn line Natural Frequency 1125 105090 Degree Out Natural Frequency 785 600Maximum Response in 750-900 Range 14 mils 3.7 mils 74% Reduction

The solution was not ideal in that the amount of mass that was added had to be limited because if too much was added, then the in-line-with-the-discharge-pipe natural frequency would be drawn too near to the 900 rpm upper limit speed. It was a bit like threading a needle. Enough mass had to be added to get the 90 degree-out-direction natural frequency below the lower speed range, but not so much that it would draw the in line with the discharge pipe natural frequency down to the higher speed value.

Desire operating range is off

response curves.

Figure 12

Vertical Pump with Mass added to reduce natural frequency

Conclusions:It is recommended that prior to adding a VFD to a vertical pump that resonance tests be

performed in the in-line-with-the-discharge-pipe direction and in the direction that is 90 degrees out from the discharge line. The very nature of vertical pumps is that they have a large amount of mass (motor) mounted on a structure with large height to width ratio. This results in natural frequencies that are often near to or below running speed. Since the damping ratios of steel are low, operating near or on one of these natural frequencies can result in very high amplitudes of vibration. Since some VFDs are able to run above the normal line frequency, natural frequencies that are above the across-line operating speed can, in some cases, also be of a concern. This paper shows a post installation solution using some rough assumptions. The results were acceptable in that it was possible to add the right amount of weight to allow the pump to operate in the required speed range without exciting either the in-line-with-the- discharge-pipe direction or the 90 degree-out from the discharge line direction. It would, however, be advisable to consider building a computer based model of a vertical pump prior to considering the installation of a variable speed drive. The model could then be utilized to test out various operating scenarios. A projected range of speeds that would be required for the analysis could be derived from the pump flow head curves and the system back pressures that would be expected from the flows needed by the process. System hydraulics and structural natural frequency considerations must be carefully considered prior to the installation of variable speed drives, especially on vertical pumps.

As can be seen from this paper, the installations of VFD drives on vertical pumps contain a number of perils that can range from electrical issues with the motor to hydraulic flow related problems as well as structural vibration issues. Designers need to take all of these factors into account when considering either the use of VFD drives on new installations or when back fitting of VFDs on presently operating systems.

3100 lbs of weight added to motor

Figure 13