Honeycomb-type wear ring increases rotodynamic pump stability

June 17, 2002
A hydrocarbon condensate pump in Oil & Natural Gas Corp. Ltd.'s Uran, India, processing plant was experiencing considerable rotodynamic instability. During many pump modifications from December 1995 to May 1999, a honeycomb-type wear ring proved to increase pump rotodynamic stability better than plain or serrated wear rings.

Based on a presentation to the 19th International Pump Users Symposium, Houston, Feb. 25-28, 2002.

A hydrocarbon condensate pump in Oil & Natural Gas Corp. Ltd.'s Uran, India, processing plant was experiencing considerable rotodynamic instability. During many pump modifications from December 1995 to May 1999, a honeycomb-type wear ring proved to increase pump rotodynamic stability better than plain or serrated wear rings.

The pump experienced the failures while handling second-stage condensate in a crude stabilizing unit. The 10-stage horizontal barrel-type pump had to handle hydrocarbon condensate, which always contains 8-9 mole % of vapor in the suction line.

Every pump that handles hydrocarbon condensate is critical because the liquid has a low potential energy compared to its vapor pressure at the pump inlet, i.e., the liquid is close to its bubble point.

Our case study shows how, given different internal pump geometries (at the sealing gaps), rotodynamic stability is the most vital variable in designing such a pump. One must understand how the pump becomes unstable when the separation level between operating frequencies and critical frequencies is low.

Background

ONGC has five reciprocating gas compressors to compress off gas produced from a crude stabilizing unit (CSU). This gas, rich in LPG and naphtha, is compressed from 1,000 mm water column (mmWC) to 15 kg/sq cm in the second stage and to 50 kg/sq cm in the third stage. The gas is then blended with the LPG as feed for further product extraction.

Each compressor has intercoolers and separators for cooling gas and separating off the condensate. Previously, the condensate from Stages 1 and 2 was sent back to the CSU. Stage 3 condensate was sent to the flare even though this liquid is rich in LPG and naphtha.

ONGC initiated a crude fractionating unit project (CFU-II) in 1995 to process this condensate along with condensate from our gas-receiving skid, which had a capacity of 130 cu m/hr. The company installed a multistage, centrifugal, horizontal, barrel-type pump to deliver second-stage condensate at 30 cu m/hr to the CFU-II unit.

Click here to enlarge image

We encountered many problems and completed remedial actions to make the pump run. The pump's piping orientation also played an important role in pump performance.

Given the composition in Table 1, the suction liquid should be about 8-9 mole % of vapor at the pumping temperature of 36° C.

Trial 1

In the first trial on Dec. 18, 1995, the pump only ran for 10 min before shutting down, too brief to yield significant observations. We were, however, able to tell that:

  • Discharge pressure did not develop.
  • Liquid temperature continuously increased.
  • Liquids were not being transferred.

ONGC found that the minimum flow line and balancing leak-off line were joined to one line, which flowed back to the suction drum. Some manufacturers do this to reduce costs, although it is not common.

Click here to enlarge image

We separated the two lines so the liquid streams flowed independently to the suction drum. A separate minimum flow continuous-recalculation valve maintained minimum continuous flow instead of using the balance leak-off flow (Fig. 1).

Trial 2

When the pump was restarted on Dec. 26, 1995, it would run for 3 hr before shutting down. We observed these abnormalities:

  • A whistling sound from the inboard seal end.
  • The non-drive end (NDE) seal was getting hot.
  • A drop in discharge pressure after 3 hr.
  • The leak-off pressure increased.
  • Vibration was as high as 24 mm/sec.
Click here to enlarge image

The whistling sound from the seal end clearly indicated insufficient seal cooling. API Plan 11 defined the inboard seal cooling system (Fig. 2). The cooling line flowed from the first-stage impeller discharge to both ends of the seal.

As part of modification, we increased the gland bore of each seal to ensure sufficient cooling at the seal faces (pressure drop across the seal flushing line reduced). We provided an orifice on the seal cooling line to the NDE seal (to have proper flow distribution for both seals). This also ensured sufficient cooling at the NDE, which is in the furthest point from the supply point of the seal flushing fluid.

Trial 3

The pump, restarted on Dec. 27, 1995, would run for 22 hr before increased vibration caused a shutdown. This time, we observed:

  • An occasional whistling sound from the seal.
  • The seal flush line became hot.
  • A decrease in discharge pressure.
  • High vibration after 22 hr online.

We dismantled the pump and found:

  • Scoring on all wear parts.
  • The shaft had run out of the coupling end by 0.23 mm.
  • A broken NDE seal face.
  • Scoring on all wear parts that developed at a single-phase rotor angle.

We reviewed the pump design and operating system design parameters to modify the pump to achieve a trouble-free running condition. From two pumps, we salvaged one shaft, a set of impellers, and one assembled rotor.

We then balanced the rotor and coupling because the coupling end had a considerable amount of run-out. Now one pump was ready for next trial run.

Trial 4

In this trial, we ran the pump several times in short-duration trial tests without finding any improvement in pump performance and observed these abnormalities:

  • A still-persistent whistling sound.
  • High vibration.
  • No discharge pressure developed.
  • Leak-off pressure increased.
  • NDE bearing temperature increased.

We again opened up the pump to the balance drum due to high vibration. Clearance between the balance drum and throttle bush increased by 1.0 mm.

We decided to dismantle the pump again. This time we found the rotor runout was 0.2 mm at the center. We again repaired one pump.

Click here to enlarge image

While investigating the pumping orientation, we decided to reroute the balance leak-off line to join at the suction drum inlet. It was previously connected to the drum close to the suction line draw (Fig. 3).

Trial 5

We ran the pump again but with no improvement. Observations were the same as the previous trials. We suspected some basic design problem related to rotodynamic stability. We asked the original equipment manufacturer (OEM) to perform a detailed analysis for rotodynamic stability because the company agreed that there could be some problem in this regard. Vibration occurred at every trial.

Rotodynamic analysis

Pumps handling incompressible fluids exhibit many interactions between rotating and stationary parts. The OEM used a computer program specially designed to handle these interactions.

In a multistage pump, there is usually a high degree of stiffening and damping. Yet this particular pump revealed conditions permitting resonance and even self-excited vibrations. The computer program analyzes these phenomena; we used it to analyze the pump simultaneously in two perpendicular plains.

Interactions between stationary and rotating elements were calculated using a matrix (stiffness, damping, and mass-coefficient). We fixed the operating speed at 2,980 rpm for all the calculations.

Click here to enlarge image

Table 2 shows the stiffness of the bearing bracket under different bracket loads, and Table 3 shows analysis results for various internal geometries. Internal geometries indicate internal profiles at the sealing gaps; i.e., at balance to throttle bush, impeller wear rings, and interstage bushings. Calculated Eigenfrequencies were not critical speeds in a true sense, which is vibration frequency equal to rotation frequency.

Objective

Our objective was to see how the rotodynamic stability is disturbed when we lowered the margin between operating frequency and critical frequency. We also wanted to figure out the best profile of sealing gaps for optimal rotor stability.

Because the bearing brackets were suspected to have contributed to damage on the original pump, we varied the stiffness parameters (Table 2). Stiffnesses between 1 and 10 million N/m had a large influence on Eigenfrequencies.

Critical speed varied from 59.4 to 60.6 cps, in line with the other analyses. The Eigenfrequencies varied from 31.9 to 68.0 cps, which was too wide and did not correspond to the model shape. Between stiffnesses of 10 and 10,000 million N/m, we found that relevant vibration influenced the frequency and that damping resistance improved.

Click here to enlarge image

Table 3 shows the damping effect with different internal configurations at the pump sealing gaps. The optimum geometry is indicated by the coefficient of damping resistance, which increases pump rotodynamic stability.

In Case 2, we reduced the clearance at the suction eye from 0.5 mm to 0.3 mm and found that damping increased from 2.2% to 5.7% with a higher Eigenfrequency.

In Case 3, we doubled the clearances from Case 2. Negative damping essentially means that there was no damping.

Case 4 maintained the same textured surfaces and a clearance level of 1.0 mm at all sealing gaps. The damping level worsened. In this case, resonance leads to absolute rotor instability.

The model analysis revealed that, if Case 2 were the new design, damping resistance would gradually decrease to zero if the clearances increased after a certain amount of rotation. This type of profile was not acceptable.

The analysis also revealed that different combinations of conventional wear rings at the sealing gaps did not yield the required amount of rotor stability at the enhanced clearance.

Case 5 included a new type of textured surface (honeycomb), which the OEM used for the first time (Fig. 4). We initially used this honeycomb surface at every sealing gap. A computer simulation showed that the wear rings worked efficiently with a narrow gap. We found that the damping and Eigenfrequencies increased considerably.

Case 6 doubled the clearances of Case 5 to simulate the wear and tear of operating the pump. The damping effect decreased to 0.3%, which was low despite instability in pump performance.

Case 6 showed that the honeycomb surface could not be adapted to all sealing gaps. So in Case 7, we only used it at the impeller eye. We again found that the damping effect increased appreciably.

We suspected that the balance drum might have a destabilizing effect. We gradually reduced the balance drum length up to 50%. With the Case 7 sealing gap profile and reduced balance drum length, we found a beneficial increase in damping. This was the best geometry of all the cases.

In Case 9 we enhanced the sealing gaps to 0.5 mm. We found the damping effect was still quite good; i.e., the pump could give good performance even with increased clearance.

Trial 6

In January 1998, we salvaged one pump and installed honeycomb wear rings. We decided to make a trial run with the honeycomb wear ring only at the impeller eye with a 0.1-mm clearance, and an interstate bushing clearance of 0.3 mm. We maintained the balance drum clearance at 0.56 mm.

We did not modify the balance drum as per the rotodynamic analysis. The pump initially ran well, but after 65 hr, pump vibration again started increasing. This incidence indicated that excitation was mainly from the balance drum.

The recommended profile after analyzing Trial 6 was:

  • Honeycomb wear ring as a stationary part at the impeller eye.
  • Serrated profile at the balance drum sealing gap.
  • Serrated profile at the interstage bushing.
  • Existing balance drum length reduced 50%.

Trial 7

After the rotodynamic analysis, we adopted every recommended internal profile except the clearance between the balance drum and throttle bush, which we maintained at 0.56 mm. We made some major modification after critically reviewing the entire history of failure and analysis. We:

  • Reduced the clearance between the balance drum and throttle bush to 0.32 mm because the leak-off flow decreased from 12 to 6 cu m/hr.
  • Refabricated all the impellers to full diameter.
  • Changed the shaft material to AISI-410.
  • Provided a groove for the throttle bushing.
  • Made the balance piston surface plain and used stelite material.
  • Replaced the NDE bearing from 7309BG to 7310BG.

Increasing shaft diameter at the NDE and changing to a harder shaft material helped stabilize the pump. We increased sealing gap stiffness by using stelite on the surface, which further enhanced wear resistance and maintained clearance well within the 0.5-mm limit.

After the many modifications, we started the last trial in March 1999. The pump continues to run with no unexpected shutdowns.

Study outcome

The original pump supplied by the OEM had stable rotodynamic behavior with water. But when we ran the pump with the process liquid, the critical speed separation margin decreased (about 5%) due to low viscosity and low density.

The amplification factor increased due to the low separation margin, which resulted in high wear at all the closed clearances. The critical speed matched the operating speed and resulted in high vibrations and fast wear, which were observed in the various trials. Other modification of the auxiliary systems, therefore, did not yield any improvements in pump operations.

Honeycomb wear ring

Click here to enlarge image

This is a special type of annular wear ring (Fig. 4) developed by OEM especially for this pump. It successfully worked as a stabilizing element. In the centrifugal pump, the pumped medium can squeeze through locations with a cylindrical annular clearance. To minimize the resulting flow losses and loss of efficiency, these clearances are kept as small as possible.

The main feature of the honeycomb wear ring was that it acts as a supplementary bearing and can significantly improve the rotor's dynamic behavior even if the pump sucks in some vapor.

The regular pattern of recesses (cells) with an axis perpendicular to the flow direction reduces the axial and circulating flow components. This leads to better rotor stabilization (compared to surfaces with circumferential grooves) and reduced losses (compared to smooth clearances).

Honeycomb advantages include:

  • Higher critical speeds and improved rotor stability vs. wear rings with circumfential grooves.
  • Lower clearance losses while enhancing pump efficiency vs. wear rings with smooth surfaces.
  • Reduced risk from contact between rotating and stationary part due to good anti-seizure properties of cylindrical cell-formed surfaces.
  • A straightforward retrofitting for high-pressure pumps.

Acknowledgment

We convey our sincere gratitude to M.L. Panwar, head of the Uran plant, A. M. Khan, head of engineering services, Sham Lal, head of mechanical maintenance, and Y. B. Tayde, sectional head, who all gave us full support to prepare and publish the findings from the Uran project, Oil & Natural Gas Corp.

The authors

Click here to enlarge image

Goutam Sardar is a maintenance engineer for Oil & Natural Gas Corp. Ltd., Mumbai, India, where he has worked for the past 15 years. He is involved with troubleshooting of rotating equipment. He has a BS in mechanical engineering, 1984.

Click here to enlarge image

Nilesh M. Daxini is a maintenance engineer for Oil & Natural Gas Corp. Ltd., Mumbai, India, where he has worked for the past 13 years. One of his duties includes troubleshooting of rotating equipment. He has a BS in mechanical engineering, 1989.