CAN-PUMP MAINTENANCE PROBLEMS SOLVED BY ENGINEERED MODIFICATIONS

Rush E. Allen
Rush Engineering Inc.
Orange, Calif.

Luey C. Chang
Bechtel Corp.
Norwalk, Calif.

Engineered solutions for vertical can-pump maintenance problems reduced maintenance costs and high production losses, and eliminated replacement costs.

The vertical can pumps shown in Figs. 1 and 2 were installed in a major waterflood operation in California. Ten vertical pumping systems were installed at three locations in the field.

Each system included a booster pump and an injection pump connected in series and mounted on a common concrete support base. Source-well water pressure was increased from less than 100 psig to 3,000 psig by the two-pump system.

During the first 4 1/2 years of operating the pumping systems, excessive failures were experienced. Motor bearing failures, seal failures, throatbushing failures, and deep erosion of the pump shaft in the throat-bushing area were presenting a nightmare maintenance problem.

Average motor thrust bearing life was only 2-4 weeks. Seals and throat bushings were failing every 4 weeks.

Fig. 3 illustrates a typical throat-bushing failure. These failures led to replacement of the pump shafts as well as the bushings due to severe scoring of the shafts.

Unsuccessful attempts were made during the first 4 1/2 years of operation to reduce or eliminate the failures. The attempts included:

• Balancing the motor armature

• Realigning the motor and pump shafts

• Adding special orifices to control seal pressures

• Modifying couplings

• Stiffening motor stands.

Standard maintenance vibration control procedures were also ineffective for reducing the failures.

Because of the high cost of production losses, maintenance crews were assigned to recondition the pumps on a priority basis. Because the standard techniques available to the maintenance department were ineffective in reducing the failures, an engineering approach was required.

The engineering approach included evaluating the complete construction concept to determine if the vertical can-pump should be abandoned.

DESIGN REVIEW

The vertical can-pump is a complex dynamic structure. Fig. 4 shows its major components. Nonrotating components are marked to the left of the assembly, and rotating components are marked to the right.

Components such as the motor stand, discharge head, motor shaft, and pump shaft represent stiffness or spring-type elements.

Components such as the stator frame, armature, coupling, and impellers represent inertia or mass elements. Other components such as the pump can, support base, and bowl assemblies are distributed parameter elements.

The relative magnitude of the stiffness and mass-type components establishes the dynamic characteristics of the machinery. The dynamic characteristics determine critical speeds or frequencies of the machinery. The critical speeds can significantly effect operating loads of the machinery.

In addition, the operation of two similar machines mounted to a common concrete base of modest size has the possibility of energy transfer between the booster pump and the injection pump.

A review of the machinery indicated that the machine components had been chosen from existing 1,800 rpm designs. To meet the pumping requirements of the waterflood application, the machinery was modified for operation at 3,600 rpm. However, only changes in pressure, flow, horsepower, seal design, and static load conditions were evaluated for the modification. As a result, the dynamic problems at higher speed operation were not anticipated.

INITIAL CONDITIONS

Vibration surveys were performed on all twenty of the pumping machines to determine if common conditions existed. The surveys were performed weekly for 7 weeks to establish trends and significant behavior from a vibration point of view.

The surveys indicated that the only significant vibration was occurring at the shaft speed, 3,600 rpm. Motion at the motor top bearing varied from a maximum of 0.91 ips peak amplitude (4.83 mils peak to peak) to a minimum of 0.08 ips peak (0.43 mils peak to peak) at the shaft speed.

Because motor-bearing maintenance requirements were directly related to high vibration, little maintenance was required on the machines with the lowest vibration.

Shaft vibration was found to vary from 1 mil to 15 mils peak to peak, and this had a direct effect on seal and throat-bushing failures. Quite often the pumps were required to be pulled out of the well for maintenance.

VIBRATION ANALYSIS

Because balancing and alignment procedures had been ineffective at failure reduction, a modal analysis approach was chosen. The technique involves the use of computer-based test instruments to define the operating deflection characteristics of the machinery.

Fig. 5 illustrates a single pumping system above the ground. The left pump shows the nondeflected geometry. The right pump shows the undeflected and deflected geometry of the injection pump during the start transient.

When power was applied to the pump, the electric energy produced a torquing pulsation. As the speed of the rotor increased, the frequency of the torque pulsing energy decreased.

During the spin up, the motor assembly on the pump stand was excited at its torsional resonance frequency (60 hz), with an equivalent motor stand torque of approximately 15,000 ft-lb as shown. This high transient frame torque had the effect of knocking the machine out of alignment after one or two starts.

Fig. 6 illustrates the typical operating deflection shapes for the pumps. The left structure indicates dominant motion at the upper motor bearing. The right structure indicates nearly equal but opposite motions at the top and bottom motor-bearing locations.

The actual motions consisted of nearly circular orbital deflections. The orbital nature of the vibration suggested dynamic unbalance with resonance amplification.

A nonoperating modal survey was performed on one pumping system. The objective of the survey was to define the resonance characteristics of the installation and determine the potential for amplification and cross feed of vibration. Table 1 presents the nonoperating resonance frequencies determined.

Modes 1 and 5 are illustrated in the left and right structures of Fig. 6. Mode 4 is illustrated in the right structure of Fig. 5. Fifteen of the 20 pumps were exhibiting upper motor orbital motion indicative of the first reed frequency behavior.

Five of the pumps were exhibiting upper and lower motor orbital motion similar to the second reed-resonance mode. The machines with the greatest vibration were exhibiting orbital motion similar to the second reed-resonance mode.

In addition, the nonoperating modal analysis revealed that cross coupling between two pumps on a common base was less than 5%. This indicated that the construction of new bases was not a requirement, and the solution to the problem should be found within each machine.

DESIGN ANALYSIS

Rather than performing elaborate instrumented tests to measure the shaft motions within the pump, analytical models were set up to evaluate the effects of design changes. The models were prepared without having to take pumps off-line to install proximity probes and other sensors.

Because the deflection analysis had revealed that individual machines were subject only to self-generated forces, separate analytical models were developed to evaluate the boost and injection pumps.

The design analysis was performed using finite-element models. Finite-element modeling is a method of defining a complex mechanical structure as a set of simple springs and masses. The individual springs and masses in the model are interconnected so as to approximate the actual structure being analyzed.

The actual structure was composed of a nearly infinite set of masses and springs; however, they were sectioned into major components as shown in Fig. 4. More complex components such as motor stator and pump-bowl assemblies were analyzed as simple beams or plates.

The result was a finite set of elements for which equations were written. The finite-element model was then used for changing design parameters to establish the sensitivity of structural behavior to modifications.

Several models were developed to establish static and dynamic characteristics of major components. A pump-can Weldment model indicated insignificant involvement of the pump-can assembly in the machinery vibration.

A pump rotor and bowl-assembly model revealed that the first three pump-rotor bending modes, and the second pump-cantilever modes, were very dependent upon guide-bushing radial stiffness.

The pump-shaft bending modes were strongly coupled to motor rocking, which had been experienced in the oil field.

The motor stand and discharge-head model was used to develop a simplified stiffness characteristic for the motor stand and discharge head. The model had no mass elements, and therefore no resonance frequencies. The model resulted in a reasonably accurate but simple beam model for the motor stand and discharge-head stiffness characteristics.

The motor-assembly model included a very detailed description of the motor frame with a simplified stiffness description of the motor stand and discharge head. The armature, coupling, and a 2-ft section of the pump shaft were modeled as beam elements.

Fig. 7 illustrates three lateral resonance modes determined by the motor-assembly model. These modes had excellent agreement with data from the model-analysis survey modes presented in Table 1.

The center mode shape of Fig. 7 illustrates the motor assembly model for the stub-shaft cantilever resonance. This behavior indicated significant potential for guide-bushing wear.

The composite pumping system model was developed using the results of the motor stand and discharge-head model and the motor-assembly model. The composite model included enough detail to simulate changes to the guide bushings and the throat bushing.

The validity of the model was supported by the prior models. However, the first three shaft-bending modes could not be verified, due to the lack of operating test data. The nonoperating test data showed error in the analytical model of +20.8 to -35.1% for the first three shaft-bending modes (Table 2).

The nonoperating test data were probably affected by bushing location and shaft eccentricity when stationary. Regardless of the unverified status, the model could be used for parametric evaluations.

Fig. 8 illustrates five modes for the composite pumping system model. The first deflection shape represents the first reed mode of the machinery. The second deflection shape represents the first pump-bowl-assembly cantilever mode. The last three deflection shapes of Fig. 8 represent the first three bending modes of the pump shaft.

When excited by rotation which is synchronous with the mode frequency, the deflection shape represents a cross section through a whirling envelope. Under synchronous vibration, the shaft does not experience oscillating stress, but rather constant stress which would be equivalent to the static stress due to the deflection shape. Imagine a jump rope held at each node and whirled around the centerline.

Under steady-state operating conditions, the shaft would experience only synchronous loads. Should some foreign matter be ingested by the pump, then a transient pulse would excite the nonsynchronous resonances.

After a reasonably short time, the nonsynchronous resonance behavior would dampen out. If the nonsynchronous resonance behavior were very near the rotating speed (synchronous), then high amplification of the transient behavior would occur.

If the resonance behavior were exactly at the rotational speed, then a whirl would grow to such amplitude that some nonlinear behavior would occur. At the pump shaft, the nonlinear behavior occurred when the throat bushing was overloaded.

Note the similarity of the right two mode shapes in Fig. 8 to the center mode shape of the stub-shaft model in Fig. 7.

High loads would be generated by the displaced mass of the whirling shape. The most significant mass on the pump shaft was the coupling.

If the motor shaft had been more substantial, then the coupling mass would have been controlled by the motor shaft. However, the motor shaft was relatively flexible, and the coupling mass was restrained by the seal and throat bushing, which failed under the loads.

MODIFICATIONS

The analytical technique indicated that the initial transient loads could be better controlled by increasing the torsional rigidity of the motor stand.

A motor stand design was chosen which included twice the steel thickness.

Diameters and flange dimensions were identical with the original design. The motor interface flange was modified to accept eight bolts instead of four to reduce the stress concentration during the start transient.

In addition to the increased motor stand stiffness, the throat bushing stiffness was increased. After some iterations, an increase in throat bushing stiffness from 4,000 lb/in. to 80,000 lb/in. was chosen.

Fig. 9 illustrates the effect of the stiffer throat bushing on one of the resonant deflection shapes. The left deflection shape shows the 4,000 lb/in. bushing model has maximum motion at the throat bushing. The right deflection shape indicates that the 80,000 lb/in. throat bushing substantially reduces the relative deflections at the throat bushing and the motor shaft. This is the desired behavior required to prevent bushing and seal damage.

The analytical results of Table 2 indicate three shaft resonance modes within 10% of the operating speed. All of these modes have the potential for excessive motion of the shaft at the throat bushing. By increasing the radial stiffness of the throat bushing, it was possible to reduce the effect of coupling unbalance on shaft motion. The right deflection shape of Fig. 10 indicates that the 80,000 lb/in. throat bushing would decrease the shaft deflection at the bushing by a factor of eight.

THROAT-BUSHING DESIGN

Based upon the parametric evaluations, a prototype bushing was designed and tested. The bushing was operated for 6 months without seal or bushing failures. The successful prototype bushing led to the redesign of the throat bushing.

Fig. 11 presents the final throat bushing design. The design had to perform several functions, including increasing the bushing radial stiffness, allowing lubrication flow, and pressure reduction at the seals. The bushing stiffness was increased by adding a collar to the pump shaft to double the effective rotating diameter.

The collar was designed so that it could be placed on a pump shaft in the field. A locking ring was loosely held to the bottom of the collar by three screws. Three long bolts passed through the collar which could be tightened from above using a hex wrench. The collar was positioned by a special plate (not shown) that rested on the discharge head, which had bolts that were screwed into the top of the collar. When properly positioned, the collar was locked into position by tightening the bolts passing through the collar to the locking ring.

The bushing housing with a labyrinth seal was installed as a unit. The labyrinth seal provided for two-stage pressure reduction and had greater clearance than the collar bushing to avoid shaft contact with the labyrinth.

A standard high-speed seal assembly was then secured to the top of the throatbushing assembly to prevent leakage of fluid from the pump. Installation and removal of the complete assembly could be performed without removing the motor.

BALANCE PROCEDURE

Initial tests with the prototype bushing were hindered by residual imbalance of the motor armature. As a result, a special balance procedure was implemented to reduce residual imbalance.

Field balance attempts using balance planes external to the bearing locations resulted in flexure of the motor shaft. The motor shaft flexure had the effect of offsetting the center of mass of the armature in the same direction as the unbalance.

Fig. 9 illustrates the effect of flexure of the motor shaft during field balance attempts. If weight were added at 180' from the armature "heavy spot" above the thrust bearing, or below the guide bearing, the motor shaft would flex as shown.

The flexure would cause additional offset of the armature mass and therefore greater unbalance. Addition of mass at O from the heavy spot outside the bearings did not remedy the unbalance by an opposite bend, rather it resulted in greater effective unbalance.

Addition of balance weights at the top and bottom of the armature core assembly directly compensated for armature unbalance, without flexure of the shaft. This required removing the rotor to add weights.

In the initial balance attempts, it became clear that the standard manufacturing clearances for bearing and flange registers was excessive for the allowable residual imbalance. The motor balance problem for the "flexible" rotor and motor frame required modification of the manufacturer's clearances for bearing and motor flange registers.

The motor-stator registers, the thrust-bearing register, and the guide-bearing register clearances were tightened from 1 to 2 mils to 0.5 mils. This required adding material to the registers, and remachining the diameters. This was necessary to minimize the effects of disassembly and reassembly during the balance procedure.

After remachining, the armature was balanced at two planes within the bearing locations. All 20 motors were "precision" balanced in the motor shop.

FIELD RESULTS

The solutions that were finally implemented had the following successes:

• Overall motor vibration reduced from 3 mils to less than 0.6 mils peak to peak.

• Monthly pump-seal failures were eliminated.

• Monthly throat-bushing failures were eliminated.

• Seal pressure-balancing orifices were eliminated.

• Motor thrust-bearing life was increased by a factor of ten.

The use of the "engineered" approach to machinery modification represented 2-5% of the cost of the replacement facilities, with even greater savings in lost production.

Copyright 1991 Oil & Gas Journal. All Rights Reserved.