– By Ron Brook –
If you are responsible for maintaining motors, you may have sent units off to be reconditioned, only to see them demonstrate unreliable performance once put back in service.
The job of customer in-house reliability technicians is to insure the reliability of any and all assets. Any asset that is returned to service and does not meet OEM specifications will be flagged and removed from service. If they have documented their work, the repair facility will have data from all final testing that should prove the asset met all acceptable specifications. Both answers cannot be correct. Or can they? This can result in accusations and endless finger pointing. The purpose of this paper is to demonstrate what could have caused this dilemma and provide the end user with the work plan to isolate and eliminate the reliability issues. This path is the only way to attack this problem for two reasons:
- The repair facility will be found to have acted in good faith and returned the asset to a condition where it will meet all acceptance criteria.
- The owners of the asset will not only identify the problem, but also have a clear answer to address the issue and return the asset to full-time service with maximum reliability.
Assets might not meet high reliability expectations for many reasons, temperature, vibration, noise, etc. Temperature readings are straight forward. The asset either operates below the maximum or not. There is not the need for higher technical expertise. Noise is similar, but noise is sometimes accompanied by vibration, but it can also be a separate problem with no connection to a noise issue.
This paper discusses the vibration issue of reliability and the test techniques necessary to ascertain why the asset does not meet specifications and how to correct this situation.
Vibration data that includes a peak average coastdown spectrum will help isolate your problem. These readings will clearly identify any areas where the resulting high vibration is due to resonance (resonance is the condition where a natural frequency is close to—and thereby excited by—a forcing function, such as one times operating speed due to residual imbalance). Impact data, when properly collected, can also identify natural frequencies. Impacting the asset to ascertain a natural frequency requires attention to several issues.
- The impact hammer must have sufficient mass to excite the structure. For example, a blade on an impeller can be tested with a small 1 lb hammer. A turbine generator outboard bearing may require a 50 lb sledge.
- The durometer of the tip on the hammer will ‘filter’ where the energy from the impact manifests itself. A very soft tip is required to center the energy in the range of an asset operating in the 20 Hz to 60 Hz region. A steel hammer tip on the small hammer is required to center the energy in the 1000 Hz to 2000 Hz range.
For this paper, the concentration will be on large rotating assets. These require 5 lb or 50 lb hammers with very soft durometer tips.
Also, if the asset has sleeve bearings, not roller or ball bearings, the shaft must be rotating slowly (250 rpm) so the shaft journal is up on the oil wedge. Failure to do this will couple to shaft into the end brackets, adding stiffness and yielding incorrect results.
The manufacturing facility can request the 2D FEA modeling results from the OEM. The majority of the problems discussed here are the result of the asset not having the same foundation as before removal. If the OEM cannot provide the FEA results and the end customer does not have this capability in-house, there are numerous consulting firms that can provide the answers with nothing more than a simplified diagram of the asset rotor. This means all rotating elements. For an induction motor, for example, this would include the shaft, and all shrunk on elements, laminations, end rings, material properties, etc.
Figure 1 shows a model of a 1000 HP 2-pole motor (the laminations, bars and end rings are included as added mass, polar and transverse inertias. The 2D model requires a support stiffness that most closely approximates the structure supporting the asset—a rotating asset bolted to a 2-inch sole plate, for example, embedded in grout that’s part of a 16-inch-thick substrate. The asset, when initially installed, should have had sufficient inertial mass (10 times the total rotating mass), and stiffness to ensure any rigid body modes are sufficiently separated from any major forcing functions, usually one times and two times operating speed. Additionally concrete must have been poured properly, with two feet of crushed stone underneath, for an equivalent support stiffness of approximately 400,000 lb/inch.
Assets are sometimes inside buildings on upper floors. These assets call for a completely different mounting scheme. Typically, the entire rotating assembly of the asset is mounted on a rigid I-beam construction. This structure is then mounted on top of spring isolation systems. This guarantees that the first two rigid body modes are well below operating speed. This process will not only move the two rigid body modes away from major forcing functions, but has the added benefit of isolating system vibration motion from the building. This is usually a necessity with floors above and below the asset are personnel offices.
When we talk about lateral modes, remember that there are rigid body modes and bending modes. The asset design engineer can predict the bending modes since they are predicated on shaft stiffness and rotor mass. Most machinery has bending modes higher than twice operating speed. There are, however, machines that operate above the first and even the second bending mode (steam and gas turbines, for example). Some large asset frames operate above the first bending mode. These issues must be efficiently addressed by the engineering contractors and building construction contractors to assure the asset rigid body modes don’t end up near those major forcing functions, like 1x and 2x operating speed.
Figures 2 and 3 show the rigid body modes, with no bending in the shaft (you can see from the mode response that the bearing phases will be in phase for the first mode and out of phase for the second).
Figure 4 shows the first bending mode. You can see its frequency is 21,556 cpm or more than three times operating speed asset operating at 3600 RPM. You can also see that the mode shape curve crosses the center line, indicating true bending. The displacement at the bearings is out of phase for this mode.
The critical speed map results in Figure 5 are the key to understanding why a returned, reconditioned rotating asset is now vibrating on its foundation. Note that a base stiffness somewhere between 50,000 lbs/inch and 100,000 lbs/inch is ideal for this motor. If the foundation is such that the equivalent support stiffness is close to 30,000 lbs/inch, then the second rigid body mode of the asset would be excited and the resulting vibration severe (this particular assets first rigid body mode is excited by an equivalent support stiffness near 200,000 lbs/inch; the red 1X running speed line coincides with the first rigid body mode curve). Equivalent support stiffness is the total stiffness in series from the bearing down to the earth. The weakest link in that stiffness chain determines the equivalent support stiffness.
Our example also illustrates that no matter what you change in a typical support stiffness setting, the first bending mode will be unchanged. The NEMA standard for testing rotating assets on isolation pads is based on understanding the critical speed map. The isolation pads lower the equivalent support stiffness to the 1000-10,000 lbs/inch range, moving both rigid body modes well below running speed. The deflection of the isolation pad must be to specification—too soft or too hard and asset could be in a rigid body mode on a test stand with very high vibration results. The specification calls for a 10% deflection of the isolation pads due to motor weight, which must be verified to prevent excessive vibration on the test stand.
A review of the peak hold averaged spectrum is necessary to identify natural frequencies that might be close to operating speed. Figure 6 is an example of a coastdown on a large (in excess of 10 tons) asset in all three axes on both bearings. One times operational speed of the asset is 3585 rpm when the data averaging began, and technicians then cut the power. The processor saved every amplitude data value during the coastdown until the process was stopped at approximately 20 Hertz, or 1200 RPM. Note that the amplitudes in three of the locations actually increase as the speed dips below 60 Hz. This identifies the residual imbalance exciting a natural frequency and driving it into resonance.
Figure 7 is an example taken on a large fan in the HVAC system of a 55-story building (20% of the building would be affected by the loss of this fan). The fan operated over a speed range as low as 200 rpm to 405 rpm. The peak average plot in Figure 7 captured the vibration over the entire operating range, revealing a resonance condition in the fan bearing pedestal near 307.5 rpm. These fans were originally fixed-speed units. Process engineers wanted to increase efficiency and save HVAC dollars, added variable frequency drives to the systems. This resulted in the potential to have the asset running speed operate directly on the natural frequency. This resulted in very high vibration that once identified by the coast down average plot, was easily rectified by ‘locking out’ this small speed range in the variable drive.
How do we measure support stiffness? Again, this may be a capability above and beyond the in-house asset reliability technicians and may require an outside consulting. The measurement requires a two-channel digital signal processor that has the capability to calculate the frequency response function (FRF- the ratio of the response motion to the input force excitation). The data is captured as impact data. Most digital signal processors will save all the data necessary to give the results plotted in Figures 8 and 9.
The measurement in Figure 8 was recorded on a large DC driven air handler with a soft support system. The first cursor has been placed on the running speed of the asset (34 Hertz, or 2040 RPM). Figure 8 identifies a natural frequency response at 34 Hertz.
Figure 9 is the dynamic stiffness plot of the same impact test. Most dual channel digital signal processors calculate dynamic stiffness by taking the FRF, double integrating the response and then taking the reciprocal. Most FRF data is taken in accelerance (response acceleration in proportion to the impact energy). The reciprocal of acceleration is apparent stiffness. Remember, it is important to keep the transducer units in order during this process. In this example, acceleration units of inch/sec/sec yields stiffness in lb/inch.
Note that the red trace (horizontal) dips at the running speed peak. This is because the stiffness is reduced at a natural frequency, thereby resulting in an increase in the vibration level. The response in the FRF will be a maximum peak and a valley will appear in the dynamic stiffness plot at the same frequency. The stiffness at this natural frequency is 9,536 lb/f (in/sec^2).
One more example: A customer recently sent a 2500 horsepower asset to a shop for reconditioning. Following the reconditioning work, repair technicians tested the asset at full voltage on their test stand. The test stand foundation consisted of a 6-inch-thick steel base imbedded in a sixteen-inch concrete floor. The motor was set directly on the steel base with toe clamps grabbing the feet and securing them to pre-drilled holes in the steel base. The measured vibration levels on this motor were all well within satisfactory limits and bearing temperatures also stabilized well below stated limits.
The shop sent the asset back to the customer, but shortly thereafter, they received a call from the customer stating that their asset reliability personnel had flagged the asset for vibration levels that were not within tolerance (the asset was not even coupled to the driven equipment). The end user asset reliability personnel did not have the capabilities to record a peak averaged plot or impact analysis. The asset was returned to the reconditioner as unacceptable.
Testing at the shop again revealed the asset was good. This time, the shop decided to send their field service technician to the customer along with the asset. After several hours on site, it became obvious what was wrong. The driver’s feet ran the entire length of the housing, front to back. The customer had only placed shims at the four corners. The field service tech shimmed the center of each leg and asset vibration dropped to levels that were acceptable to the asset reliability personnel. Costly? Absolutely. Avoidable? Yes. A learning experience? Hopefully.
So the next time your reconditioned asset will not meet reliability standards, and you have eliminated unbalance, misalignment, soft foot and every other possibility—take a look at your support stiffness. What is changed? It could be something as simple as shorter shims instead of shims designed to support the entire asset foot. It could be corroded shims that do not supply full surface support when restacked. Foundations do not last forever. You may want to check for cracks in your concrete foundation or vibration in concrete pads, as this would signify that the fill below the concrete has washed away. The point is you need a complete evaluation of the operating environment, which means taking more data. If the motor had no problems when it came out, but does after reinstallation, it may not be the reconditioning work that is at fault.
Remember—additional test data and inspection may be required to isolate the problem. The correct data can isolate the problems and help you resolve these issues before cracks appear in your relationship with the service provider.
About the Author
 Ron Brook is Supervisor of Engineering Services, Emeritus for Integrated Power Services. Ron has 49 years of vibration experience. His work covers vibration analysis, modal analysis and operating defection shapes, rotor dynamics modeling, vibration isolation and damping, finite element analysis and root cause failure analysis. Ron can be reached at 215-805-2006 or rwbrook@ips.us |
