While on site you start collecting and reviewing history. There are several key points that you observe right away. Before the failure, process conditions change, causing a different load condition in the turboexpander.

The original equipment manufacturer reviewed this data as well and concluded there should be no problems since the unit was operating well within the design limits.

Previous rotordynamics work indicated a small imbalance, but that was to be expected with the machine loading and unloading.

In every case there was no warning of the pending failure. This would eliminate aerodynamic excitation, build up of the impeller or degradation of the impeller. Bearing temperatures would remain within normal limits and would not exhibit any issues of concern.

The impeller was investigated, and there was no evidence of foreign particle damage or liquid carryover.

The cause of the failure was due to a rub that occurred between the impeller and the shroud. Clearances are tight in the turboexpander to obtain desired efficiencies and performance. As the impeller spins at normally very high rpm, the clearances are tighter due to the centrifugal load. During a process change it is possible for the temperatures to vary several hundred degrees. Due to the thermal lag between the housing and the impeller, it is possible for clearances to close up and lead to a failure.

An approach to avoid these failures and make sure the equipment functions properly across all operating ranges is as follows:

1. Perform a process analysis of the unit concerning all steady state and transient conditions of the turboexpander.
2. Utilize computational fluid dynamics to model the impeller to look at local temperatures and determine the possibility of any liquid formation.
3. Based on design clearances, perform a finite element heat transfer analysis of the turboexpander to evaluate the clearances.
   The model should consider the “spin up” conditions to obtain the final location and configuration of the impeller. The analysis would have to be a transient heat transfer analysis to include the thermal lag in the system.
4. It is important that the turboexpander handle all the fluid states that come from the process. In some cases there may be ice formation or dust particles that can cause issues if not considered in the design.
5. The rotordynamics of the system are important, with rotor stiffness a major consideration.
   The rotor should also be designed to handle some imbalance due to solids buildup. Remarkable thrust conditions must also be handled by a robust design.
6. A good seal design appropriate for the application must be considered. Make sure the materials are compatible to handle temperature transients as discussed in this article.
7. Develop and incorporate the necessary instrumentation to address the reliability issues of the machine.

Turboexpanders are basic in design and are not that complex in structure. However, like many pieces of rotating equipment, the details of both process and mechanical design are what affects a successful outcome in operation. A qualified professional engineer competent in rotating equipment should be involved with troubleshooting and implementation of any fix of this equipment.

For more information, visit www.knighthawk.com or call (281) 282-9200.  Click here to find suppliers of rotating equipment.

You’re offshore on one of the largest rigs in the world. It was put together with all the latest equipment with all the “bells and whistles.” The rig was built, erected and went into production in record time. As with any new unit, it takes a while to get all the “bugs” out, tune things and get lined out. While observing the rig, you see something you don’t like — process fluid dropping out of a flanged joint. The flange connection is a new quick connect type joint (QCJ). The maintenance crew chief is called and he sends someone out there to tighten the bolts on the QCJ, and the small leak stops. You’re relieved and in subsequent observations, you see no leak. A few weeks later, you walk by and you see “drip, drip, drip …” You call maintenance again, they tighten the clamp, and the leak stops. Once again, you don’t see it at first, but it pops up again in several weeks. You ask, “What’s going on?”

Your company has approved the QCJ for this service; it has been tested, and it is used widely all over the company. The QCJ comes from a well-respected manufacturer, and your company has had a long relationship with them. You ask yourself, “What is the problem” and call for help.

The situation described above is not necessarily an isolated case in the business. In cases where we have a line in a service that is cycling “hot to cold” due to the inherent design of the process, the QCJ is subjected to cyclic load and pipe stress loading. Depending on the QCJ design, typically a high seating stress or seal contact stress is required to prevent leakage. It is important to torque the QCJ properly in accordance with the manufacturers’ specifications.

With all that said, in most cases the external pipe stress loading will derate the allowable pressure for the QCJ. Pipe stress loading must be included as part of the design. The combination of pressure and external loadings is usually too complicated to calculate. Most applicable codes and standards handle the pressure-only case quite well, and even the manufacturers own design tools can do a good job. But when high external loadings are imparted onto the QCJ, the ball game becomes more challenging. Fortunately, the finite element tool can be used to look at these QCJs. The methodology is as follows:

1. Lay out the pipe design.
2. Perform a code pipe stress analysis.
3. Define the loads for the QCJ.
4. Specify the type of QCJ you want.
5. Develop an FE model of the QCJ.
6. Analyze the loads.
7. Specify a large QCJ or move the clamp to a strategic location.

The problem of external loads on QCJ is challenging in some applications, but it is also a problem with standard flange joints as well. Many times our leaks in the field are misdiagnosed as a gasket problem or whatever, when it is simply the external pipe stress loading and the cycling thermal load causing the problem. It is recommended that a registered professional engineer competent in this field reviews any designs. As with any piece of hardware implemented in industry, the complete ball game has to be defined and analyzed prior to it being put into service.

Share

A progressing, or progressive cavity pump consists of a single helix metal rotor that turns inside a double helix rubber stator. This forms a series of pockets that traverse the length of the pump, gently pushing fluid from the suction to the discharge in a smooth, pulseless stream.

Progressing cavity pumps are unmatched in their ability to transport viscous, abrasive, and shear sensitive fluids. They can pump exotic solutions such as shrimp and brine without crushing or tearing the product. They routinely handle sewerage sludge, ceramic slurry, wax, chemicals, confectionery ingredients, grinding compound, sand, cement, grout, putty, pulp, ground meat, pigments, glue, paste, grease, paint, lime slurry, jelly and soap.

Their precise internal geometry and minimal back flow make them accurate metering pumps, eliminating the need for a flow meter. You can deduce volumetric flow rate by adding a tachometer pickup to any rotating part of the pump and motor assembly.
To operate reliably, however, these pumps must stay within certain operating limits. Since they are a true positive displacement pump, discharge pressure theoretically spikes to infinity if the discharge is blocked. In reality, the motor stalls under these conditions, but usually not until after the flange bolts stretch or discharge piping bursts.

To minimize back flow, there is an interference fit between the metal rotor and the rubber stator, so these parts rub against each other during normal operation. The pump depends on the fluid stream to carry away the heat generated by friction. If the pump runs dry, the stator overheats until the rubber melts or scorches. At this point the ruined stator must be replaced.

These pumps handle difficult fluids that make it particularly challenging to monitor the process to prevent this.

Protecting against damage from over pressure requires the addition of a pressure switch. Pressure gauges and switches are prone to clogging or plugging when pumping the kinds of fluids usually handled by this type of pump. Protecting gauges and pressure switches with conventional diaphragm seals buys some time, but these devices become plugged themselves within a few days of operation.

The most successful approach to date has been to use an annular ring seal, usually referred to as an isolator ring.

An isolator ring consists of a rubber ‘inner tube’ clamped in a steel ring. The assembly fits between flanges in the process pipe. Clear instrument oil behind the rubber membrane transmits pressure to the gauge or pressure switch. The motion of the process fluid continuously cleans the inside of the ring assembly.

Protecting the pump against damage from run dry conditions is trickier; several methods have are used, each with advantages and disadvantages.

FLOW METER:An obvious method is to install a flow meter coupled to a signal relay. If flow falls below a certain threshold, electrical contacts open, stopping the pump. This is straight foreword and reliable, albeit expensive. The only flow meters that can handle the fluids typically conveyed by progressing cavity pumps are magnetic or sonic meters, which cluster near the high end of the cost scale.

THERMAL MASS DISPERSION: This device monitors flow by measuring the rate of heat dissipation from a probe in the flow stream. Inside the probe are two
thermocouples and a heating element. One thermocouple is next to the heating element, the other is some distance away.

An electronic circuit monitors the difference in the temperature of the two thermocouples. With no fluid motion around the probe, the thermocouple closest to the heating element registers a higher temperature than the reference thermocouple. When fluid flows past the probe assembly, it dissipates heat from the ‘hot’ thermocouple, so temperature at this thermocouple drops. By measuring the difference in temperature at these two points, the electronic circuit determines if there is sufficient flow present.

The thermal mass dispersion device is more cost effective than conventional flow meters, but is subject to certain limitations.

Fluids that build up a coating can insulate the probe, degrading sensitivity. Highly abrasive fluids can erode the probe to the point where it no longer functions.

These devices can be difficult to calibrate because when a progressing cavity pump runs out of liquid, it becomes an air compressor! Since thermal dispersion devices measure air flow as well as liquid flow, configuration and calibration can be difficult. For example,
you can’t test this device by simply turning the pump off; you have to force the pump to run dry to produce an air stream in the pipe to check if it is properly detecting fluid loss. This places the stator at risk.

CAPACITIVE DETECTORS: Depending on pipe size, cost is usually higher than a thermal mass dispersion device. These are installed on the suction side of the pump.
They do not monitor flow; they monitor the presence or absence of fluid in the pipe.

The device consists of an insulated ring sandwiched between flanges on the pump inlet. The principle element in this device is a metal plate lining the inner circumference of the ring. A layer of electrical insulation separates this plate from the process fluid. The metal plate functions as one side of a capacitor circuit; the fluid in the pipe functions as the other plate.

With fluid in the suction pipe, capacitance of the circuit is high. If the pipe is empty, capacitance of the circuit is low. An oscillator applies a reversing polarity signal to the plate. The circuitry measures the dampening effect of the capacitive plate and infers the presence or absence of fluid within the boundaries of the ring.

A coating or build-up on the inner surface of the device affects the reading, but a simple re-calibration compensates fro a buildup to 0.250″ or more without adverse effects. They are immune to false reading caused by air flow during run dry conditions because they do not detect flow, they detect the presence or absence of material within the boundary of the device.

THERMAL MONITOR: Another run-dry protection device is a thermocouple monitor buried in the stator. A small hole is drilled through the stator tube and a thermocouple is inserted. As the rotor turns inside the pump, it rubs across the thermocouple for a portion of each revolution, allowing the thermocouple to sense the temperature of the rotor. A signal relay amplifies and monitors the output from the thermocouple. If the pump runs dry, friction heats the rotor and stator. The thermocouple detects this temperature rise and triggers a signal to stop the pump.
The advantage to this approach is low cost. However, fluid temperature fluctuations can trigger false alarms. Conversely, cold ambient temperatures can delay response.

The hole involved prohibits the use of this device in sanitary, food, or high purity applications.

Different fluids will affect response time. Non-lubricating fluid such as lime slurry will elicit a fairly fast response, but fluids which exhibit higher lubricity such as polymer flocculent may take several hours to evaporate. The resulting gradual temperature rise may take several hours to reach the set point on the alarm relay. By that time the stator may have suffered substantial damage.

POWER MONITORS infer the presence or absence of flow by noting changes in electric power by the pump motor. However, the power dip caused by run dry conditions is short lived and difficult to capture. The reason is that running dry causes rapid heat build up. The fluid film lubricating the interface between the stator and rotor quickly evaporates, so friction increases within seconds after loss of fluid. Soon, he heat build up in the stator can causes the rubber to revert, making it gummy, and further increasing friction.

PRESSURE MONITORS
Pressure monitors use pressure to infer flow conditions. This method is simple and straight forward with the advantage that you can monitor high pressure with the same device.

Positive displacement pumps should always be fitted with over pressure protection. If operated against a closed discharge (dead-headed) a positive displacement pump builds up pressure until a system component fails. The motor fuses may blow, but
it is equally probable that a pipe fitting will burst.

Pressure switches are more reliable on slurries than pressure relief valves and easier to reset than rupture disks.

Several manufacturers offer control boxes to interface with the pressure switch and isolator ring assembly.

These boxes incorporate local push button control stations and include a timer to allow the pump time to prime itself each operating cycle. Some of these boxes incorporate seal flush controls as well. To use this approach, it is necessary to understand the difference between static pressure, friction pressure, and total pressure.

Static pressure results from the pipe being filled with liquid and is present even when the pump is idle. It is not influenced by pipe size, number of
fittings, or viscosity. Static pressure is determined solely by fluid density and difference in height between the pressure switch and the outlet of the pipe.

In the example in figure 6, the outlet of the discharge pipe is 12 feet above the gauge, so static head is 12 feet, which exerts 5 psi of pressure.

Friction pressure results from the flow of liquid through a pipe and is present only when the pump is running. It depends on flow rate, size
and length of pipe, number of fittings, and fluid viscosity.

Total pressure is the combination of static and friction pressure. This pressure is observed directly by reading the gauge on the pump discharge.

When the pump is idle, the gauge shows static pressure. When the pump is operating with flow present the gauge shows total pressure. In the example in figure 7, total pressure is 25 psi.

For run dry protection, set the low pressureswitch midway between the static and total pressure. In our example the correct setting for the low pressure switch is 15 psi.

When the pump is running correctly the low pressure switch signals that flow is present. If the pump runs dry and flow stops, the pressure falls back to the static pressure. This causes the low pressure switch to trip.

Flow Detection with Pressure Switches.
Suction or Discharge?

Is it better to monitor pressure at the inlet or the outlet of the pump?

Let’s use an example to study pressure changes at both points in the system.

Figure 8 shows a typical flooded suction and pump arrangement.

At the beginning of the cycle, the feed tank is full so the suction gauge reads positive 10 psi. As the pump lowers level in the feed tank, the reading on the suction gauge gradually drops.

As long as the pump sustains flow the discharge gauge reads the total of static and dynamic friction head, which
is +25 psi.

This pressure represents 10 psi of elevation head combined with 15 psi dynamic friction head.

Figure 9 shows the same pump after the level falls to the middle of the feed tank.
Now the suction gauge reads zero pressure, because feed tank level is at the same elevation as the suction gauge.

Lift to the outlet tank and pipe friction hasn’t changed, so the discharge gauge still reads +25 psi.

.

.

.

Figure 10 shows the situation when the feed tank level is almost to the bottom of the feed tank. Suction pressure drops into the vacuum range, but the pump is still operating perfectly, lifting fluid into the suction port.

Notice that the discharge pressure holds steady at +25 psi.

Figure 11 shows the end of the cycle.

The feed tank is empty, so now suction pressure is atmospheric.

The suction gauge changed direction and went back up to zero!

.

There is still liquid in the discharge pipe, but no flow through the pump, so discharge pressure drops to the static
head of the discharge pipe, + 10 psi.

Notice that the suction gauge shows the same pressure at tank empty as it did at tank half full.

There is no way to discriminate between these two conditions by reading the suction gauge.

Where would you set a pressure switch on the suction side of the pump?

Even if the feed tank is higher than the pump, it is difficult to monitor flow on the suction side.

The pressure also goes up past zero when the pump turns off, so where would you set a pressure switch on the inlet manifold?

Even with a perfect set up, monitoring suction pressure is tricky. Figure 13 shows an ‘ideal’ pumping system. Pump suction is flooded and suction pressure never falls below atmospheric. However, pressure switches are subject to “dead band”.

.

Dead band is the change in pressure required to operate the switch and make the contacts transfer. In most switches, this dead band is about 1.5psi. That translates into ≈4 feet of
head.

In other words the best accuracy you can expect for a switch on the suction side of a pump is an error range of about 4 feet elevation under ideal conditions. It’s difficult to make generalizations, but dynamic friction head on the discharge side usually ranges
anywhere from 5 to 150 psi.

Setting a discharge pressure switch in the middle of this range provides simple clear cut indication of flow.

There are systems with low flow rates and low pipe friction, where there is insufficient friction head to trip a pressure switch. Your options are to find or create some pipe friction. If there is a check valve on the pump discharge, there is usually enough friction pressure upstream of the check valve to operate a pressure switch.

In a system with minimal pipe friction, it is easy to create some.

Adding a reducer nozzle at the terminus of the system as shown in figure 14 usually creates at least 5 to 10 psi pressure differential between flow and no-flow conditions.

A minimum differential of 5 psi is sufficient to dependably trip a pressure switch designed to monitor flow.

Protecting progressing cavity pumps from over pressure and run-dry conditions makes economic sense. A pump protection device usually costs less than the replacement cost of one stator, and avoids the cost of an unscheduled shut down to service the pump.

For safety reasons, a pressure switch or similar device should always be used with any positive displacement pump. The different methods of protecting progressing cavity pumps include:

• Flow Meter

• Thermal Mass Dispersion Device

• Capacitive Detector

• Thermo-couple Monitor

• Electric Power Monitor

• Pressure Switch

A pressure switch is a simple economical solution to monitoring progressing cavity pumps. Applying a pressure switch to monitor pump conditions requires an understanding of some basic fundamentals of hydraulic behavior.

Here are a list of approved suppliers of progressive cavity pumps.

Share

Onyx Isolator Rings have traditionally been supplied with mechanical gauges and mechanical switches to monitor process pressure and pump performance. These do a fine job, but many users are changing over to solid state switches. This paper will discuss some of the advantages of each.

Principle of operation:

Mechanical pressure switches operate with either a piston and spring, or a bourdon tube and mercury vial. In the piston and spring arrangement, when pressure builds up to the point where it crosses the set point, the piston pushes on a snap acting switch to transfer mechanical contacts.

In a mechanical bourdon tube switch, a bourdon tube tilts a rocking beam fitted with a glass vial of mercury. When pressure builds up to the set point, the beam rocks over and mercury in the glass tube rolls from one end to the other, where it submerges electrical contacts completing the circuit.

In contrast, a solid state switch uses a metal diaphragm about the size of a dime with a micro-miniature strain gauge etched onto its dry surface. Increasing pressure changes the impedance of the strain gauge. A digital comparator monitors the output from the strain gauge and compares it to the user specified set point. When the set point is reached, the electronic circuit turns on a solid state relay, completing the circuit.

Accuracy: A typical mechanical Bourdon tube pressure switch has a published repeatability of ±1% compared to the A-B 836 series which has a published repeatability of ±0.2%, a five-fold improvement in performance.

Dead-band adjustment: A Bourdon tube switch with a range of 2 to 60 psi has a minimum dead-band of 3 psi. The A-B switch with the same range has a minimum Dead-Band of 0.30 psi, a ten-fold improvement in performance.

The difference is even more striking in the low range. A Bourdon tube mechanical switch in a compound range has a minimum dead-band of 1 psi compared to the A-B switch in the same range which has a minimum dead-band of 0.1 psi. If you are trying to monitor suction pressure on a pump, this is the difference between best resolution with a mechanical switch of 2.3 feet of water column and the solid state switch with a resolution at 3 inches of water column.

There’s a big difference in field adjustment setting as well. Mechanical switches typically require screw drivers or wrenches, and setting the switch – even with a socalled “dial” – is a trial and error method. In contrast, the solid state switch is a matter of pushing a few buttons and programming in the exact set point you want.

Attitude: Snap acting mechanical switches can be mounted in any position, but mechanical switches with mercury elements MUST be mounted vertical and upright or they will not function. All solid state switches can be mounted in any position without affecting their operation.

Over Pressure: Mechanical switches with a bourdon tube can not be subject to pressure over the maximum range. In contrast a solid state switch can tolerate a substantial over pressure without adverse effects. An A-B switch in a 60 psi range has a 160 psi maximum working pressure; a 150 psi range solid state switch can go to 400 psi without ill effects.

Price: The cost of a mechanical switch by itself is generally lower than a solid state switch; however, the solid state switch has a digital read out for pressure, so the gauge is superfluous. A solid state switch is (almost) always less expensive than a mechanical switch + gauge assembly.

Power requirements: There are three advantages to the mechanical switch:

1. You don’t need to know the operating voltage in advance. A mechanical switch can work on either AC or DC power at almost any voltage. In contrast solid state switches come in two distinct flavors: Low voltage (12-30 volt DC) and high voltage (90 to 250 volts AC) so you have to know what power level you are dealing with before placing an order.

2. Mechanical switches can handle higher electrical loads. The Bourdon tube element example can handle 4 Amps at 120 volts AC, compared to the solid state switch which can handle 2.5 Amps at the same voltage. Solid state DC switches are generally rated at ¼ Amp at 24 VDC.

3. Mechanical Switches use 2-wires, whereas a solid state switch needs a third wire for a common.

Capillary tubing introduces three effects on the performance of pressure-sensing instruments:

1. Temperature effects. Temperature changes cause the liquid inside the capillary tube to expand and contract, changing the volume of the fill fluid. The resulting error is a function of the total volume of the tubing, pressure instrument, and isolator ring. Because the rubber sleeve in the isolator ring has a much lower modulus of elasticity compared to a diaphragm seal, it can absorb most of the volumetric change resulting from temperature differences throughout the usable temperature range for isolator rings. A typical error in gauge reading through a temperature swing from 0°F to 120°F is about ½ psi depending on isolator ring size and gauge type. This is roughly a quarter the error expected with a standard 60 mm stainless steel diaphragm seal.

2. Elevation effects. As you change the elevation of the gauge with respect to the isolator ring, you introduce an elevation error. This error is due to the static pressure of the liquid in the capillary tube. The change in gauge reading caused by elevation changes of the pressure sensing instrument can be calculated in advance using the following equation:

P {psi} P Elevation{feet} sp gr actual gage = − ÷ 2.31∗

Our standard fill fluid is a silicone oil with a specific gravity = 0.967 at 77°F.

Observe polarity: If the gauge is above the isolator ring, then the elevation term in the above equation is positive; if the gauge is below the isolator ring, the elevation term is negative. If the gauge or transmitter has a zero adjust capability, the elevation error can be eliminated completely by re-setting the zero adjust to compensate for the elevation change.

3. Response time. Capillary tubing introduces a response time lag in the instrument reading. This delayed reaction time is influenced by:
• Length of the capillary tube
• Internal diameter of the capillary tube
• Control volume of the pressure-sensing instrument
• Viscosity of the fill fluid, including temperature effects fill fluid viscosity
The control volume of a pressure-sensing instrument such as a gauge or transmitter is defined as the change in volume required to deflect the bourdon tube or sensing diaphragm from zero to 100% reading. The smaller the control volume, the better the performance. Instruments with smaller control volumes exhibit less temperature error and time lag than instruments with greater volume. As a general rule, higher range instruments have a smaller control volume than lower range instruments; for example a 100-psi gauge has a much smaller control volume than a 15-psi gauge. Another general rule is that electronically amplified devices have a smaller control volume than mechanical devices; again, to use an example, an electronic transmitter has about 1/100th the control volume of a bourdon tube gauge of the same range.

Onyx isolator rings are supplied with capillary tubes with ID = 0.075 inches, and the fill fluid is Silicone with a viscosity = 100 centistokes. Typical time lag with 5 feet of capillary and a 4 ½” turret case gauge with a 60 psi range is about 2 seconds.

4. For those users who want a more exact prediction of the performance of a pressure measuring system consisting of an isolator ring, capillary tube and pressure sensing element, you can bring the full capability of modern mathematics to bear on the problem:

Technically:

Where:

Err = Error in reading expressed in inches H2O water column.

E = Expansion factor. If you use degrees F and inches for all the terms in the above equation, then E for Silicone fluid = 0.000600

Rs = Spring rate. For isolator rings, Rs is related to nominal size such that:

RS = 100/Nom size, in

The error as a percent of reading is therefore:

Error%= Err/ Measured Span in inches H20 * 100

Volume of Iso Rings:

Article written by Cliff Knight of KnightHawk Engineering.

Once again, a debottleneck project is finished, and the plant is complete and coming on line. As it turns out, Process Engineering identified ways to achieve more pounds per hour by upgrading a quench tower with new trays and by adding a couple of new nozzles in the tower.

The turnaround went smoothly, and testing went well. One of the key components in the system was a turbine compressor train pushing cracked gas through the system. Things are looking good and, as the plant rotating equipment engineer, you have made your rounds and all the equipment is within allowable ranges regarding temperatures and vibration. You go home with a good feeling and rest easy as you get ready for the next day when the plant is slated to reach full production and start making product.

While the quench tower modifications were easy, the major part of the project involved the compressor upgrades that consisted mainly of tilt pad bearings, impellers and controls. New recycle lines, knockout drums and so forth were installed to handle the added capacity. It was not anticipated that there would be any problems, as this unit had previously run well in all aspects.

Compressor drive train noise and vibration

However, things went south when the train went on recycle. The noise was louder than you’ve ever heard, and operations reported cracking in one of the compressor drums. To make matters worse, the pedestal bearing between the turbine and compressor had unacceptable vibration levels. Since the train is critical to plant operation and the plant was unable to run, your office became a good stop over point for management to vent.

The problems that occurred with this upgrade are not unusual. The turbine and compressor were outfitted with the latest hardware as a technology upgrade in addition to the capacity increase with the change out of the impellers. The plant was “bit” by two problems with this upgrade.

Acoustical Analysis

First, the reconfiguration of the recycle system along with the impeller upgrade excited the fourth acoustical mode in the compressor recycle piping, which coupled to the drum natural frequency and led to the failure. No acoustical analysis of the system was done. The drum diameter was unnecessarily large and could easily couple into any acoustic pulsation.

The second problem was that no sensitivity study was done on the bearing stiffness when considering the new tilt pad bearing in the turbine. The pedestal bearing was too weak, and there was an interference at the first mode that caused the pedestal to vibrate. Another part of the problem was that the bearing base plates were not rebuilt during the upgrade.

Some points to consider when upgrading a turbine compressor train such as this one are as follows:

1. Perform a complete process analysis of the new train.

2. Develop a new process specification sheet for the equipment.

3. Revisit the old rotor dynamics studies of the existing equipment and operation.

Determine that the old simulation properly identified the critical components in the system. It would also be a good idea to revisit all previous analysis with the latest rotor dynamics tools. Finally, develop a tuned or normalized model of the existing operating system.

4. Develop a finite element model of the pedestal bearing to determine actual stiffness.

5. Perform a complete rotor dynamics study of the system.

6. Perform an acoustical analysis of the recycle piping and discharge. Make sure there is no interference with blade or vane pass in the system.

7. Conduct a review of results with process, maintenance, mechanical and operations before signoff.

This is just a brief review of the ballgame. Turbine compressor trains are complex, and each has its own characteristics. All work should be reviewed and approved by a professional engineer that is competent in rotating equipment.

Source: http://www.knighthawk.com/Related%20Pages/062_BICSept08.pdf

Article written by Cliff Knight of KnightHawk Engineering.

Today is your first day at your company’s polymers division, and you are amazed at how small everything seems. You were just transferred from a world-scale ethylene plant as some manager up the food chain wants to diversify your career. Inside, you are a little disappointed because you feel you will miss the love/hate relationship that you had with the big gas turbines and “cracked gas” compressors. You always knew that the ethylene plant provided product for the polymer plant, but you never really focused on what the downstream plants did with it.

Polymer Gear Pump Problem

At the start of the morning meeting, someone started talking about the polymer gear pump that fed the pelletizer. From what you could tell, it sounded like a piece of “farm equipment” that someone put in the middle of the process. Your ears perked up as you heard that the bearings in this pump were polymer lubricated. Furthermore, you understood at this meeting that the bearing operated at 650 F. You then observed the production supervisor’s stress level go up when the production engineer reported a rise in bearing temperature despite running the same Melt Index and turning at the same rpm. 

As the new reliability engineer at the plant, the production supervisor told you to investigate. You wondered to yourself what all of the “big hurrah” was about. Then your buddy told you that all of the product generated at the ethylene plant is fed to this plant, which then uses it to make double the profit.

The plant has four trains and this one was in jeopardy with the higher temperatures. The bearings have failed three times over the past year. The company is now being adversely affected with customer relations and with losses in profit.

When a world-class polymers plant goes down due to an unplanned outage from a polymer gear pump, a production facility can lose millions of dollars. As it turns out, that “farm equipment” is one of the most complex pieces of equipment to analyze. Polymer lubricated bearings are non-Newtonian. In other words, the viscosity will vary as a function of pump speed and bearing temperature for a given polymer.

Most gear pumps and other rotating equipment do not have a high variance of viscosity except by temperature.

To further complicate things, the bearing load is affected by the discharge pressure and the discharge pressure is affected by the speed of the gear pump, which in turn affects the viscosity, which affects the load carrying capability. Oh yes, everything is tied together. This problem is a fully coupled heat and mass transport balance with viscous dissipation. To make matters even worse, the load distribution across the bearing is also affected by shaft deflection.

Modeling a Polymer Bearing

To fully model a polymer bearing requires an iterative methodology between structural and fluid/thermal analysis. A general procedure is as follows:

1. Establish acceptance criteria for the gear pump bearing design.

2. Develop a fluid/thermal model of the bearing to determine the response.

3. Perform a finite element analysis of the pump case and shaft to evaluate deflections.

4. Iterate the solution until the correct force balance is obtained and the energy balance is satisfied.

Based on the results, the bearing design can be modified or changed to meet the design constraints for the bearing and polymer. Many production facilities desire to increase production to meet demand requirements. Typically the gear pump output limits the plant production and profitability of the unit. Depending on the gear pump design, small design internals can be changed to increase the output of these units. In addition to an increase in output, reliability can also be increased.

Historically, the gear pumps have been under designed for the applications and have led to rate limitations at production facilities. With the advancements in numerical technology and algorithms to address the physics involved with these pumps, design and reliability issues can be properly addressed. These pumps typically affect the profitability of a facility as much as a large gas turbine or a compressor.

These units are driven by complex physics and must be designed and addressed using advanced numerical technology to yield the best results. As always have a professional engineer competent in gear pump design involved with any modifications or failure analysis.

Source: http://www.knighthawk.com/Related%20Pages/BicNov08.pdf

Most of us have been in a situation at a production or manufacturing facility where we have asked ourselves the question, “Why was this equipment specified for this service?” Problems with the equipment have caused you to delay your vacation or had you out at the plant on the weekend. Your boss keeps asking you, “If we have all this technology and resources, why can’t we fix it?” The equipment manufacturer has been in business for 100 years and has a good reputation, so you think, “We must be doing something wrong or there is a design problem.” You appreciate all the responses from the equipment manufacturer and they have a good relationship with your company. But the fact still remains, it does not work.

Well, maybe the answer is simple. There is a mismatch between the equipment specified and the process. Typically, a major part of a product’s research goes into the process design and not the static and rotating equipment used for the process. A process development usually starts in the lab. Later, a pilot plant is developed and the process is proved out. The process looks good, the economics look good and management says to build a pilot plant.

Pilot vs Production Plant

However, there is one problem; the pilot plant only had to run a week at most. A production plant needs to run at least a year without a shutdown. The process developers put together a set of piping and instrument diagrams and a process flow sheet so the engineering and construction contractor can design and build the plant. Typically, a parade of vendors will perform all their sales pitches and the best equipment is chosen for the process application. There is a lot of pressure to get the plant built and running and start producing the product. Rightfully so, management is also aware that you can only “hammer” on a plant design so long and there will still be unforeseen problems that need to be worked out. The plant starts up and immediately a piece of equipment becomes infamous due to its lack of performance.

Take a step back and ask yourself, “What is the best equipment for the application?” It probably does not exist. You do not have time for your mechanical group to develop a new piece of equipment.

Conceive the Right Machine for Your Application

Well, maybe the right decision is the “Mean Machine.” The Mean Machine is simply a “tweaked” or modified form of a particular vendor’s piece of equipment. It frequently contains components from other vendors that normally do not conduct business together. But in the end, it works. The process of developing the Mean Machine begins with an analysis of the required application of the existing equipment and its associated problems.

The goal is to put existing technologies together to develop the equipment that best fits the application. Usually, an engineering group capable of analysis and design of equipment can perform and coordinate the design efforts.

Case study

A process production facility required a piece of equipment to grind its product for preprocessing. This was not anticipated and was discovered in the middle of a startup. Without this the plant could not run. The application was high temperature and high pressure. Grinders were available for the process but none at high pressure and high temperature. Pressure vessel shops were available to build containment equipment. The solution was to marry the grinder application with the vessel and machine shop. The grinder was successfully completed in days instead of 28 plus weeks for a complete independent design.

The Mean Machine does not always have to be a complex research and development project. Look at putting technologies and companies together to fix your problem. One of the greatest aspects of American heavy industry is its ingenuity, innovation and “out of the box” solutions to complex problems that demand high resources in a short time.

An implementation policy that should be considered is as follows:

1. Define the requirements for the device.

2. Create process and mechanical specifications.

3. If an existing device is in service, analyze and determine the limitations and problems.

4. Conceptualize using “out of the box” thinking for the new design.

5. Evaluate the process and mechanical response.

6. Perform risk assessment of the new design/HAZOP.

7. Design a detailed design.

As always, the designs should be reviewed by a professional engineer competent in machine design.

Source: http://www.knighthawk.com/Related%20Pages/BicJan09.pdf

Problem Definition

It is the middle of the night and once again the phone rings at your home. Before you answer it, you know what it is about.  You are the maintenance manager at a major ethylene plant and before you left work today vibration levels started to climb on the pedestal bearing on the turbine compressor train. It happens several times a year, and no one really knows why. You have brought in the original equipment manufacturer (OEM) and all sorts of consultants to study the problem because this has shut the plant down at least three times in your tenure as maintenance manager. The “phenomenon” theories are free flowing and you have listened to opinions how operations are setting up a transient affecting the balance of the plant.  So you head out to the plant and once again it is shut down due to high vibration in this bearing.

Having coffee at midnight with one of the 35-year operators who has lived with this equipment brought great insight.  He told you that the problem did not start until the train was sped up by 300 rpm to increase production. He also told you the problem occurs with weather changes. It just so happened a “Blue Northerner” came though the previous afternoon and the temperatures dropped 30 F.

Now it just so happens that the structure is elevated and the turbine compressor train is on the second floor. In the mean time you have to start the plant back up and you do what you always do to get it to run — you change the oil temperature to stabilize the bearing. As the night wears on you look at the rotordynamics reports from the past few years. All of the work indicates a stiffness in the pedestal bearing but none show where it comes from. The next day you investigate and discover that the stiffness does not include the elevated structure or consider the ambient thermal growth of the structure. At this point it has all come together for you.

Consider the structural stiffness of pedestal bearings

Sometimes a turbine train will consist of a turbine and compressors all in one unit. A pedestal bearing is installed in some cases. Rotordynamics studies are conducted to ensure the unit will operate as intended.

However, many boundary conditions are assumed for the studies that may not always be valid. For example, the structural stiffness of pedestal bearings. Two problems exist with pedestal bearings regardless of the specific bearing design itself — the stiffness can affect the critical speeds calculated and the pedestal itself can experience vibration problems as well.

One important parameter in the rotordynamics study is the stiffness of the pedestal in all-principal directions. Of course the bearing damping coefficients and bearing type are important as well. The stiffness of the pedestal bearing is not always easy to obtain. Sometimes the support structure should be figured into the ball game as well. The structural stiffness can be calculated using a finite element tool. A structural dynamics model can be developed to determine the dynamic response of the support structure. The boundary conditions can be derived from the finite element model and incorporated into a rotordynamics model. If problems exist, the pedestal model can be revised as required to obtain the desired response.

How to tackle pedestal bearing problems

Pedestal bearings are a major player in many turbine compressor and turbine generator trains. A suggested procedure for tackling the issue is as follows:

1. If the unit is existing, perform a field study to determine natural frequencies of the structure.

2. Create a base rotordynamics model of the train.

3. Perform a sensitivity study on the stiffness of the pedestal bearing.

4. Develop a finite element model of the equipment’s support structure.

5. Incorporate the stiffness into the rotordynamics model. This should include any temperature effects.

6. Develop a detailed rotordynamics evaluation including the support structure.

7. Look at changes in bearing design and structure that can detune the system away from detrimental criticals.

Every situation is unique to itself and each application should be reviewed by a professional engineer who is competent in rotordynamics.

Source: http://www.knighthawk.com/Related%20Pages/BICFeb09.pdf

It is your first week on the job in a plant that is old but new to you. You find out that, once again, the plant is faced with another failure of an integrally geared compressor.  To make matters worse it has failed five times in the past two years. This last failure occurred within just a few weeks of the previous failure. Now the plant manager has had enough and profits of the plant are suffering. One of the main functions of your new job is to perform a failure analysis of the latest failure and to assist with getting the compressor train operating reliably.

Bolt Failure

The failure has always been in a tie bolt that fastens the open face impeller to the shaft. This failure consisted of the third stage impeller separating from the shaft while running full speed during normal steady state conditions. The steady state failure is new for you because you have been told by the operators that the other failures have occurred during a start-up or shutdown.

To start the project you read all the previous work on the other five failures. You also see where the compressor manufacturer has been called in on every failure. Every failure has been attributed to corrosion fatigue. The words “corrosion fatigue” catch your eye and you recall that in your 20 years of experience, rotating equipment OEMs (original equipment manufacturer) have almost always attributed failures to corrosion fatigue, slugs or surge. You rarely if ever found an admission of a design error as a probable cause from an OEM. What really caught your interest on this project was that the corrosion seemed to always occur on the same impeller bolt. 

Root Cause Failure Analysis

Realizing that there must be something else going on, you have a contractor perform a root cause analysis of the failure and the findings are interesting. The contractor concluded that the cause of the failure was due to an inadequate design of the stage impeller bolt and fastener system. Finite element analysis (FEA) indicates high stress in the failed area.

This provides no allowance for any normal dynamic stress that impeller components would experience. During the investigation a few other items came to the table. Contributing factors were excitation of the impeller from a process instability while running coincident with blade pass frequency. Also maintenance procedures for bolt-up regarding mount and dismount cycle life of the tie bolt were an issue.

In most cases the problem was easily fixed with a change of material for the tie bolt, illumination of the process instability and scalloping the impeller to change the natural frequency. The problem was addressed from all fronts to ensure that the unit would not fail again.  It just happens that on initial tightening most of the load of the tie bolt is carried with only the first few threads of the tie bolt. If tightening continues such that the first three threads yield, the next three threads start to pick up more load. If the bolt is over-torqued, it is possible to compromise the bolt preload if a significant number of threads have yielded. With a loss of preload the full dynamic load is transferred to the tie bolt and a fatigue failure of the tie bolt could occur.

Open Face Impeller Issues

In approaching open face impeller issues to avoid failures, consider the following:

1. Perform a CFD (computational fluid dynamics) analysis of the gas path to insure a sound aerodynamic design.  Consider any secondary wake interaction or acoustic effects.
2. Perform a detailed structural dynamics analysis of the impeller and tie bolt assembly. FEA is a good tool to evaluate the design.
3. Look at natural frequency interferences with the Campbell and nodal diameter mode shape interference diagrams.
4. If necessary create a Goodman Diagram for any anticipated dynamic loads.
5. After the impellers are built, perform “ring checks” to evaluate the natural frequencies. The dynamic stresses can be checked with strain gauges mounted on the impeller. Mode shapes can also be determined and evaluated.
6. Compare the measured frequencies to the FEA results for the impeller.
7. When the impellers are installed, the impellers can be instrumented to look at vibration during the first runs.

Not all of these steps are required but the return on investment is quite high for most large machines. All work should be performed under the direction of a professional rotating equipment engineer competent to do this work.