Case of the Whirling Pump This article is a vibration case study using Operating Deflection Shape (ODS) and Experimental Modal Analysis (EMA) tools and computer animation techniques to solve difficult rotating machinery problems. Mr. Tim Schuyler and Mr. Steve Werness from Xcel Energy Fort St. Vrain Generating Station requested the services of Full Spectrum Diagnostics to perform a vibration investigation on a vertical river water supply pump whose vibration had significantly increased recently for unknown reasons. The Fort St. Vrain facility is located in Platteville, Colorado, and was both that state’s only nuclear power plant, and the United States only commercial High Temperature Gas Cooled (HTGC) reactor design. The plant was decommissioned in 1989 and converted from a nuclear fueled facility to a natural gas fueled steam generation facility in 1996. The site added heat recovery steam generator (HRSG), and a second gas turbine in 1998. A third combustion turbine with HRSG was added in 2001, providing a current combined rated output of 720 megawatts. Two more combustion turbines are planned to supply up to 300 additional megawatts. Cantilever mounted vertical pump designs will typically have natural frequencies near operating speed points. These so-called "reed" modes are a function of the height of the cantilevered structure (above deck), the length of the pump shaft (below deck), the mass of the drive motor, and the "footprint" of the mount interface structures. Each of these parameters effect the mass and/or stiffness of the overall structure, thus the location of the natural frequencies. Also, as a note, symmetric systems (like this unit) will have two fundamental cantilever natural frequencies (N-S and E-W). These modes are fairly close together, but may appear below operating speed, above operating speed, or even straddle the running speed point. This makes identification of the modes a top priority in establishing a proper structural correction, if required. In addition to the planned Experimental Modal Analysis used to detect natural frequencies, an Operating Deflection Shape (ODS) analysis was also planned to investigate the operating characteristics. This analysis is helpful in resolving unbalance, looseness, misalignment, and bowed rotor faults. Operating analysis also provided an evaluation of historical repairs to the pump foundation on these units. Xcel Energy provided periods of both operating and non-operating test conditions for the ODS and Modal analyses, respectively. Machine Design Information Drive: Westinghouse AC Induction Motor 200 HP @ 1180 RPM Rolling Element Bearings: Driven Machine: Worthington Single Stage VTP River Pump 1180 RPM Direct Drive Coupled 3-Piece Shaft Assembly 2 Spider Bearings 6-Blade Impeller Capacity 6000 GPM, Total Head: 70 ft Even though the operating vibration problems were found only on the P4119 pump, natural frequency testing was performed on each unit for comparison purposes. Experimental Modal Analysis
Finding Hidden Machinery Faults With Operating Deflection Shape Analysis
by Daniel T. Ambre The generating station has a pair of vertical river water pumps used to control the cooling water in the facility. The P4119 and P4119S units are similar in design, both manufactured by Worthington and driven by Westinghouse 200 hp AC electric motors. Overall vibration levels on the P4119 unit were noted to have "jumped" from a stable trend of 0.150 ips to levels approaching 1.0 ips. The dominant frequency was found to be the 1x response at 1180 RPM. A resonance condition was suspected, however, a history of foundation issues suggested a deeper underlying problem or contributing factor.
An Experimental Modal Analysis is a diagnostic performed on structures to determine their natural frequencies, damping, and mode shapes. This analysis is performed with the machinery in a non-operating condition, which limits background vibration from other machinery sources. An impulse (usually a hammer blow) is applied to the structure to impart broad-band energy to the system. This energy functions to excite all of the system‘s natural frequencies, similar to the response
of a bell or tuning fork to a sharp impact. This hammer has a load cell built into the head that measures the force parted to the structure and allows this signal to be compared to the accelerometer response, via a transfer function, resulting in a measurement of the frequency response of the structure.
Given enough response locations, each system’s natural frequency can be curve fitted and its mode shape animated with specialized modal analysis software. The Mode Shape is an effective visual representation of the deformed structure when amplified by a driving force.
The natural frequency analysis of each unit revealed some curious results. The problematic unit (P4119) appeared to have significant operating margin on its fundamental natural frequency modes. The operating speed was well above each mode. In addition, the secondary unit (P4119S) had no amplified response; however it was operating smoothly at a point between its EW and NS cantilever modes, and with significantly less speed margin?
A summary is presented in Tables 1 and 2. Figures 2 & 3 show the measured natural frequencies in relation to the operating speed (cursor).
Vibration levels in the P4119 River Water Pump were found to be considerably above the vibration alarm levels typically recommended for a pump of this configuration. The historic overall vibration trend obtained from Mr. Schuyler indicated that the levels had recently increased. Operational vibration velocity levels were measured at 1.0 inch per second (ips) on the motor frame, and up to 1.5 ips on the outboard end (bell housing) of the motor. Typical alarm levels for vertical pumps would be set at roughly 0.290-0.340 ips, depending on vertical elevation. The harmonic content (multiples of 1x RPM) were considered low and acceptable.
From experience with vertical pumps of this type, an amplification of a natural frequency is typically the cause of unexplained vibration response(s). The symmetric cantilever design dictates that two natural frequencies should be present (inducing North-South, and East-West motions), i.e., the so-called "reed" modes of the pump. The biggest dilemma in correcting resonance problems is defining the locations (frequencies) of the modes. The two dominant corrective methods used to alter the natural frequency in resonant systems are the addition of a stiffness (bracing) and/or addition of mass to the physical structures. Unless the natural frequencies are first defined, the wrong correction method may be applied resulting in increased vibration levels instead of the desired reduction. With symmetric systems (like this one) the worst case is when the operating speed is between the two modes, which increases the potential of inducing additional problems no matter what corrective method is chosen.
In addition to the Experimental Modal Analysis, this analysis included Operating Deflection Shape (ODS) analysis. This test method is described below.
Operating Deflection Shape Analysis
An Operating Deflection Shape is measured with the machine at its’ normal operating condition. This analysis measures the machines’ response at a specific time or frequency. Both amplitude and phase information are collected at various locations on the structure and, via special software, the vibrating "shape" or response of the machine can be animated.
These animations show the analyst how the machine is moving during normal operation. Note that this is not a resonant response of the machine, but its operational response. The forces within the machine are responsible for the motion, or shape of motion measured with this analysis tool. For example, the unbalance response of any rotating system will produce a response or driving force at 1x RPM. Misalignment and looseness generally produce synchronous multiples of running speed (2x RPM, 3x RPM,etc.).
Examining the operating shapes gives the analyst additional insight into the root cause of many vibration related faults. Combining this analysis with an Experimental Modal Analysis is very effective in ruling-out suspected machinery problems associated with resonance.
Analysis & Conclusions
In this analysis, the P4119 pump indicated some unexpected characteristics. Typical resonance amplification will induce a linear response in the direction of, and similar to the mode shape defined in the EMA. The closer the driving force is to the natural frequency, the more mode shape resemblance in each animation.
This characteristic was not found in the ODS animations. In the P4119 pump operating shapes, a "whirling" or "orbital" ODS animation motion was noted. This visual response,along with the seemingly adequate resonant margin, was another indication that this system was unusual.
The whirling orbital motion in Figures 5 and 6 suggested that the excessive vibratory response was NOT due to resonant amplification.
ODS animations near a resonance will typically correspond to the linear mode shape. The animations suggest that the pump rotor (extended below deck) was driving the response in the motor. A possible reason for this is excessive clearances in the seals and "spider" bearings that support the rotor. Significant leakage from a problematic seal in this pump tends to support the theory that radial clearances are larger than desired.
Additional animations without the motor (Figure 6) more clearly show the whirling motion in the motor standoff support structures. The final animation in Figure 7 shows an isolated view of the support base plate structures. The foundation amplitudes suggest that vibration levels are higher than normal; however, the dense measurement mapping confirms that the plates are moving in unison (in-phase).
Out-of-phase motion between plates (grouted interfaces) or bolted joints would imply looseness response. These animations confirm that the foundation is not the source of the vibration problem. The relatively new bolted joints in the base plate are considered acceptable based on this analysis.
The final analysis performed was an "added mass trial" on P4119. Mass is added to resonant systems to alter the natural frequencies. Adding mass (with sand bags) should both reduce the natural frequency, as well as dampen its response via the shifting sand. Since the natural frequency for this unit was below operating speed, adding mass should increase the operating margin. If a stiffener support were applied instead, the likely result would be an increase in vibration since the operating margin would be reduced.
The Xcel Energy maintenance support acquired several sandbags totaling close to 120 lbs of weight. The unit was started up and the bags were stacked on the motor. No change in vibration levels were noted regardless of the amount of sand applied.
The unit was shutdown for a natural frequency (impact) test. The test indicated that the natural frequency at 1027 CPM was shifted lower by the added 120 lbs of mass to 990 CPM, or approximately 6%. If resonance were the root cause of the elevated vibration levels, a significant drop in the 1x RPM amplitude would have been noted during pump operation. This proved that resonance was not the dominant contributor to the excessive vibration levels.
Based on this analysis and discussions with Xcel Energy management and maintenance personnel, it was recommended that the unit be removed and inspected. It was suspected that some internal unseen fault was responsible for the elevated vibration levels.
Disassembly and inspections of the hardware showed that a locked coupling was likely the root cause of the excessive vibration. Continued operation in this "locked" state resulted in significant wear in the spider bearings, the shaft, and rotor bushings. This extra clearance allowed the rotor to "whirl" as a unit, inducing the orbital motion noted above deck.
The operating animations (ODS analysis)proved to be invaluable in isolating the root cause of the rotor problems.
Dan Ambre, P.E. is a Mechanical Engineer and founder of Full Spectrum Diagnostics, PLLC, a Full Service Predictive Maintenance Consulting company. Dan specializes in Resonance detection, Experimental Modal Analysis, and Operating Deflection Shape machinery diagnostics.
Full Spectrum Diagnostics provides Vibration Analysis level I, II, and III training and certification, as well as training in advanced diagnostic techniques. Dan is a certified software representative for Vibrant Technology, Inc., the creators of ME’scope VES software tools. He also provides ME’scope VES Software Training targeting the In-Plant Vibration Analyst.
Animations referenced in this article are available via email at modalguy@aol.com. Please visit www.fullspec.net
There is also an online Modal discussion at MaintenanceForums.com
Print friendly PDF below:
Download OctNov_Vibration.pdf
Hi Thomas,
What is missed by most readers of this article is that the problem was unknown by the author when he inspected the machine. A modal test was done first because that was the suspected culprit. Only when resonance was eliminated as the cause was other testing done. Since Dan had ME’scope available as a tool, it made perfect sense to continue using it to complete the diagnosis.
If you follow this link, a discussion has already been posted. In it, I made the comment that if you have a ratchet in you tool pouch, would you use it or a combination wrench.
http://maintenanceforums.com/eve/forums/a/tpc/f/3751089011/m/2751001993
Posted by: Jon McFadden | November 13, 2008 at 12:00 PM
Wow,
All that effort to say what was inevitable, replace the unit. Its not like it was a defective unit from the get go. If it is vibrating rebuild it! Cause and effect! Sometimes one must rationalize to justify ones exsistence.
Posted by: Thomas Reinhart | November 13, 2008 at 09:38 AM