Using crest factor analysis for early failure detection of progressive cavity pumps
In my 19 years of experience with vibration analysis in refinery and chemical plants and 15 years in the oil and gas sector, I have set up, managed, or supervised numerous vibration programs for equipment ranging from small pumps to large turbomachinery. However, I had no prior experience with vibration on progressive cavity (PC) pumps, which are used for pipeline transfer of water and oil.
When initially setting up a vibration monitoring program for these pumps, we used the overall vibration analysis parameter alongside ISO severity charts, which included higher vibration limits due to the field-expedient installation methods commonly employed in the oil and gas industry. However, the overall vibration analysis parameter quickly proved to be ineffective for monitoring the pumps, prompting us to seek a better indicator for assessing their operational health.
This article presents a novel approach to assessing the health of a progressive cavity pump through vibration analysis. It highlights the limitations of traditional overall vibration measurements, which proved unsuccessful in detecting failures in the pump component of the equipment train. Instead, this study identifies the crest factor—typically used in bearing fault analysis—as a more effective indicator of pump wear and impending failure in progressive cavity pumps.
Equipment setup
My plant uses progressive cavity pumps for pipeline applications involving both water and oil, in sites with either tanked or tankless systems. In tanked systems, the suction pressure ranges from 1 to 2 psi, while tankless systems typically operate around 100 psi. The pump discharge pressure ranges from 100 to 165 psi. The pumps are driven by a variable frequency drive (VFD), with speed ranges of 55-440 rpm or 63-339 rpm, depending on the system. The VFD speed is adjusted based on the suction vessel's level.
The most common failure mode in the PC pumps at this plant is loss of flow and pressure, typically caused by stator failure. Selecting the right stator material is essential, as it must be compatible with the chemical properties of the fluid being pumped and the operating pressures and temperatures. Choosing the wrong stator material can drastically shorten the stator’s lifespan.
Progressive cavity pump metrics
Table 1 below shows the percentage of failures for each type of PC pump in service (excluding all other pumps) and the mean time between repairs (MTBR) in months for each pump. Note that the water pumps have a higher MTBR compared to the oil pumps.
A key issue with these pumps is the number of starts. Tankless systems experience 10 times the number of starts per day compared to systems with tanks.
- Average starts per day for a PC pump in a tankless system: 33.0 starts
- Average starts per day for a PC pump in a system with tanks: 3.4 starts
The MTBR is very low in comparison to my previous experience at both other plants and oil and gas organizations. Not only is the low MTBR an issue, but the cost of repair for the equipment averaged $50,000 per repair, with some failures impacting production.
Table 2 below breaks down the pump failures by discipline (Automation, Electrical, and Mechanical). Most of the failures are mechanical and are broken down by the top three failure modes in Table 3.
Pump failure analysis indicates that the stator is the most common failing component, typically identified by a loss of discharge pressure and flow. Connection rod/pin joint and seal failures are the next most frequent. Most of these failures are identified through vibration routes using crest factor screening. (The crest factor is defined as the ratio of the peak to the RMS [Root Mean Square] of a time-domain signal.) However, some stator failures happened very rapidly and were not detected during the monthly routes. These rapid failures were caused by temperature and pressure excursions and by a high concentration of sand in the production separators.
Establishing a vibration program for PC pumps
The vibration route program was initiated in July 2020. Taking vibration readings on the PC pumps was a new experience for me, as this was the first time in over 28 years of reliability engineering that I had worked with them. For each pump, Figure 4 shows vibration points of concern set up across the motor, gearbox, and pump.
In the past, I typically referred to the ISO 10816-3 Vibration Severity table for analysis. However, in this situation, traditional vibration levels based on ISO standards were not effective in assessing the health of the PC pumps or in predicting when they were nearing the end of their life.
In the plant environment, the initial alarm thresholds were set at 0.07 IPS (inches/second) for the first alarm and 0.18 IPS for the second alarm. However, in the oil and gas sector, pump installations are less engineered, so I adjusted the alarm levels to 0.18 IPS for the first alarm and 0.44 IPS for the second. This adjustment was necessary; otherwise, 80% of the equipment would have triggered both the first and second alarms, resulting in an overwhelming (and unmanageable) volume of data to analyze. Increasing the vibration limits reduced excessive investigations.
Despite this, the PC pumps continued to fail between monthly vibration checks, even without triggering the first or second alarm. Upon further review of the data and the readouts from the Emerson CSI 2140 vibration analyzer, I noticed that every pump had some form of alarm in the crest factor analysis parameter. I began tracking this data for the failed equipment and adopted the crest factor severity chart from Fluke (shown in Figure 5, below), applying it to the pump's connecting rod/pin joint and stator/rotor to estimate the health of these critical pump sections. (Fluke uses the term "Crest Factor +" for bearing damage analysis, as the crest factor is particularly effective at identifying early signs of bearing failure.)
For pumps with unacceptable crest factor values (>15), a year’s worth of data was analyzed, showing that these pumps typically failed within 3.4 months or less in our tankless facilities, and within 2.4 months in facilities with tanks. The pumps in tanked facilities were expected to last longer due to the lower number of starts. Interestingly, in most cases, the overall vibration analysis parameter did not exceed the 0.44 IPS alarm threshold required to trigger action.
The pump in this example, progressive cavity (PC) pump used in water service at a facility with tanks. The pump is installed without grout on a channel base plate on an elevated steel platform that sits directly on the ground. This example will first review the pump’s data following a repair and then at the end of its life.
Figure 7 shows vibration parameters for the pump following its replacement in September 2021. All vibration levels are within acceptable limits according to field standards. All crest factor values are below 5. However, except for the motor, all other points show some level of crest factor anomalies. I observed this trend across the entire population of PC pumps. Also of note is the lack of PK-PK parameter alarms.
Figure 8 shows Spectra 1 (POH/POA/P3V) for review for the new pump. Overall vibration and crest factor vibration are within acceptable limits. The red area circled in Figure 8 highlights an interesting issue observed in some of the PC pumps. This occurred when the pump exhibited a squeak related to speed, which could be seen on the time waveform oscillating in sync with the squeak. However, I could not find a correlation between the squeak and increasing crest factor vibration or other vibration parameters on the pump. Pumps with and without the squeak both showed increased crest factor vibration over time as the pump progressed toward end of life.
A year later, the crest factor vibration had increased to the point where the pump was reported as being at the end of its life. However, the pump continued to operate for another five months before failure, surpassing the 2.4-month average time to failure once acceptable levels of crest factor vibration in the pump are exceeded. Upon reviewing the crest factor data (see Figure 9), I identified damage to the connecting rod leading into the rotor/stator, as well as some damage to the stator. In my experience, a damaged stator/rotor typically deteriorates faster than a damaged connecting rod. Since this was a tanked system (which experiences fewer starts), it makes sense that the pump lasted longer than average.
A closer examination of the spectra reveals low overall vibration levels, with some visible impact in the time waveform. A review of Spectra 2 (shown in Figure 10) shows an above-range reading of 0.135 IPS, and the zoomed-in spectra indicates the running speed of the pump (6.25 Hz at 0.12 IPS) along with two multiples. Considering the crest factor of 22.6, using overall vibration alone would have missed the issue.
Additionally, the PK-PK measurements (see Figure 11) indicate elevated levels of PK-PK vibration. While the crest factor provided an earlier identification of the problem when compared to overall vibration, and has proven to be the most effective indicator of pump wear and end of life, the PK-PK waveform analysis is another parameter I used to assess the pump once the crest factor reached Level III. Figure 11 shows a side-by-side comparison of the overall vibration (IPS), crest factor, and PK-PK waveform.
The highest overall vibration was found in the motor (MOH), not the pump, where the crest factor data indicated a failing pump. The PK-PK waveform also showed elevated vibration, similar to the crest factor, but I found that the crest factor alarm levels in Figure 5, the Crest Factor + Severity Chart, were more effective at predicting failure earlier then the PK-PK. The crest factor parameter normally goes into alarm status before the PK-PK parameter. From Figure 11 above, Table 4 below tallies which parameter was higher for each point:
During my three years of collecting data on the progressive cavity pumps, I encountered two other interesting issues. One was an oscillating squeak, which I could observe on the time waveform while taking data (previously discussed this issue, circled in red on some of the provided spectra in Figure 8). I tracked the pumps exhibiting this squeak but found no correlation between the squeak and an accelerated failure rate. I also performed a natural frequency test on five pumps with the squeak to check for resonance but did not find any natural frequency in the relevant range.
The second issue was a loud, constant squeal (100-120 dB) that was undetectable with the 2140 vibration analyzer. This squeal typically occurred with new or recently repaired pumps and disappeared within 2-3 months. I contacted the OEM about both the squeak and the squeal. While they acknowledged experience with the sounds, they could not provide an explanation or solution. My hypothesis is that the squeal was related to tight clearances between the stator and rotor, which likely disappeared after some wear-in. However, I’m unsure if this contributed to faster deterioration of the stator, as I found no supporting evidence in my data.
Vibration analysis techniques used
This section reviews the vibration analysis techniques used on progressive cavity pumps. Below is a summary of the techniques employed:
- Overall Vibration (IPS) – This parameter was not effective for analyzing the pump section, including the rotor/stator and pump connecting rod. However, it was useful for monitoring the motor and gearbox.
- PeakVue® (High-Frequency Enveloping) – PeakVue® is intended for use in detecting faults with roller bearings and gear teeth and was used for the bearings in the motors/gearbox and the gearbox itself.
- Crest Factor Analysis Parameter – This parameter was useful for evaluating the health of the rotor/stator and pump connecting rod. It provided a good indication of the pump’s end of life. However, some pumps failed between vibration checks, often due to operational issues that caused rapid stator degradation. To accurately trend and predict failures, an online monitoring system would be required, using the crest factor analysis discussed in this paper.
- Natural Frequency Testing – This testing was conducted but did not find any natural frequencies that matched the operational frequencies of the system.
- Peak-Peak Analysis Parameter – This measures the maximum peak-to-peak acceleration in the time waveform in Gs and is useful in confirming the end-of-life status in the pump and pin-joint sections. I also used this with the PeakVue® to review the condition of roller bearing assemblies and gearboxes.
- Motion Amplification Video – This tool was particularly useful in persuading project engineers to place pumps on concrete foundations. The technology magnifies the motion of the equipment, highlighting the significant difference between pumps mounted on concrete versus those on dirt. After mounting the pumps on concrete foundations, vibration levels were significantly reduced.
Machine pump health
The most effective analysis parameter for determining the health of the pumping part of the machine train was the crest factor, which proved to be far superior in predicting failures in the pumps' stator and rotor, the most frequently failing components. Overall vibration and PeakVue® were more effective for analyzing the motor and gearbox. Based on this, I set up three different reports in the CSI AMS Machinery Manager software and reviewed each route for potential issues with the progressive cavity pumps. As mentioned earlier, the MTBR for these pumps was significantly lower than what I had experienced in previous organizations.
About the Author
Craig Cotter
P.E., CMRP
Craig Cotter, P.E., CMRP, is a mechanical engineer. He has more than 30 years of experience in reliability engineering and maintenance management. Cotter has a B.S. in mechanical engineering as well as an MBA. He is a retired U.S. Army Colonel. Contact him at [email protected].
Ziad Wardeh
Ziad Wardeh is senior production engineer at Oxy.
Tony Nguyen
Tony Nguyen is a chemical engineer and facilities engineer at Oxy.