Comparing the vibration performance of identical or like machines can be used to detect defects and judge their severity.
What is Comparative Analysis?
Comparative analysis is a simple yet powerful technique used to compare the performance of identical or similar machines to detect anomalies and assess the severity of their condition. The crux of the technique is comparing measurements between like components operating in similar conditions to discern whether the components are operating satisfactorily. In layman's terms, we are performing an "Apples to Apples" comparison.
If the operating conditions are not similar, then external influences may cause changes in performance. This article focuses on applying this technique to vibration and other PdM technology measurements, but it equally applies to process parameters.
Why Use Comparative Analysis?
A maintenance engineer at a power plant is often requested to investigate components that exhibit "unusual" noises or vibration. If the engineer is lucky, the component in question is a piece of rotating equipment within the scope of the PdM program and historical data is readily available for trending purposes.
In the other 99 percent of these cases, the first questions are "What is it?" and "Where is it located?" The component may be a static structure like piping, grating or a snubber or a piece of rotating equipment like a motor, pump or fan. This technique provides the most value to components that have never received a PdM inspection. The following list provides some situations where comparative analysis is beneficial:
Limited or no historical data available to assess equipment
No applicable "standard" alarm limits or lack of confidence in existing alarms
An observation of "unusual" noise, vibration, etc.
No additional resources required-uses existing data and measurement capabilities
How is Comparative Analysis Used?
Comparative analysis can be applied to vibration data, including overall, spectral, time-waveform (TWF) and enveloping/demodulation measurements. Additionally, it can be applied to other PdM technologies and process parameters. The key point is comparing measurements from identical/similar machines operating at identical/similar conditions. Unfortunately, no standard definition for similarity exists. Common sense and engineering judgment are required to assume that the similarity condition applies.
Parameters to Consider
The following list provides a number of parameters to consider when deciding if components and conditions are similar. Realize that the list is not exhaustive, and all parameters may not be applicable to the components in question.
Flow, pressure, temperature, load, speed, force
Conditions: ambient temperature, location (indoor/outdoor), nearby machines (in-service/out-of-service), atmospheric pressure
When the equipment is operated: time of day, portion of cycle, seasonal (summer/winter), high/low temperature service, tide level (high/low)
Layout: suction/discharge piping, structure (baseplate, grouting, pipe-supports), component orientation (vertical/horizontal)
The following case history provides an example of how comparative analysis was successfully used at San Onofre Nuclear Generating Station.
The component in question is a 40-gpm, centrifugal, overhung, make-up pump, designated at 3P1019. The pump is driven by a two-pole, 7.5-hp, AC induction motor. This pump is operated on a quarterly basis, and is tested within the plant's inservice testing (IST) program for safety-related pumps.
Prior to 2007, the vibration measurements collected from this pump were limited to an overall velocity measurement and a 1-kHz spectrum due to software limitations. In 2006, a plan was developed to add high-resolution spectrums, TWF and spike-energy measurements. The baseline measurements were made in early June 2007 for this pump.
A review of the vibration data showed that the overall amplitudes were below all alert limits and the trend and spectral plots appeared normal. However, the spike-energy (Figure 1) and TWF plots (Figure 2) suggested a pump bearing problem. At this point it was necessary to confirm if the pump bearings were degraded and determine the severity of the degradation.
Figure 1. A waterfall plot comparing the baseline spike-energy spectrums from make-up pump 3P1019. The pump bearing defect frequencies, harmonics and sidebands are indicative of early bearing wear/damage.
Figure 2. Baseline time-waveform plots from 3P1019 on June 4, 2007, indicating impacting at the pump bearings.
This pump is an excellent candidate for comparative analysis due to the following conditions:
No historical data for TWF and spike-energy measurements
Two pumps per unit (four total) with similar layout/orientation and operating conditions
Parameters measured/controlled during IST: suction pressure, flow rate, discharge pressure and differential pressure
Table 1 summarizes the process parameters collected during the baseline testing of each pump. Note that the flow rates are identical and the differential pressures are within a few percent of each other.