Editor's note: Reg Christmas of Kemira Chemicals Inc. contributed to this column.
This column discusses the results of an eight-year reliability study conducted at a chemical plant. The plant uses six horizontal split-case cooling water pumps to provide cooling water to the production facility. These six pumps operate in banks of two units, taking suction from three cooling towers and feeding cooling water to three plant processes:
- Crystallizer condenser (Cooling Tower A feeding Pumps A and B)
- Process 3040 (Tower B feeding Pumps C and D)
- Process 5060 (Tower C feeding Pumps E and F)
Company management wanted to evaluate the operation of the pumps with regard to their reliability and energy consumption. The eight-year pump reliability study began in 2008 and includes the best economic use of the installed equipment. The pumps are 12x10x12XL model 8100 (11.4-inch impeller diameter), driven by 150-horsepower (hp) motors at 1,775 revolutions per minute (rpm).
As shown in Figure 1, the best efficiency point (BEP) is at 4,800 gallons per minute (gpm) for the 11.4-inch impeller. The actual flow varies from approximately 1,500 to 6,500 gpm, depending on the process demand. As the study showed, Pumps A and B operated, on average, at a higher flow than Pumps C through F.
The main indicator of reliability degradation, which typically indicates a need for repair or overhaul, is a significantly increased level of vibrations. To assess this parameter, the overall level of vibrations was measured at four locations: inboard and outboard bearing housings of the pumps and motors as well as at horizontal, vertical and axial directions at each position, equaling a total of 12 data points per monthly test. The highest value of the 12 is plotted on the charts shown in Figure 3 (page 16).
As shown by the performance curve, the pumps operate within a wide range of flow, sometimes substantially too far to the left of the BEP and sometimes too far to the right. As a result, both cavitation (high flow) and recirculation (low flow) were concerns. The system went online in the 1990s, and the present management inherited the problem and wanted to find the most economical and practical solution.
Both vibration data and flow are plotted as a function of time, as recorded in Figures 2 through 7. The time axis represents each monthly recording.
Pumps are sitting on the ground level, approximately 5 feet above the water supply level, so net positive suction head available (NPSHA) is approximately 29 feet. The performance curve in Figure 1 indicates that the pumps would occasionally run out of net positive suction head required (NPSHR) when operated beyond approximately 5,500 gpm. Among the suggestions for improvement was replacing these split-case pumps with vertical turbine pumps in order to increase NPSHA to more than 40 feet, with an impeller/bell of the vertical pump positioned about 6 feet below the water level. The cost of this change was evaluated against present maintenance.
The following observations can be made from analyzing the eight-year data shown in Figures 2 through 7:
- Pumps A and B operate, on average, (disregarding occasional abnormalities) toward higher flow as compared with Pumps C through F. Pumps A and B operate between 2,000 and 6,000 gpm. Pumps C through F operate between 2,000 and 4,000 gpm. While far from perfect, this data implies that Pumps A and B operate, on the average, closer to the BEP, so they are expected to have better reliability.
- There seems to be no clear correlation between vibration and flow, meaning there is no correlation between failure rates and flow.
The traditionally accepted vibration levels for pumps are well-known and, depending on pump size and type, are roughly below 0.20 inches per second (in/sec) (rms values) and are usually understood to apply to new installations. For the existing installations, field-recorded values are somewhat more liberal, with 0.30 in/sec being the warning level and 0.50 in/sec being the alarm level. No known studies exist, however, to correlate vibration to failures because of the time such studies would require to accumulate reasonable statistical data to quantify the theory. The eight-year study discussed in this column is among the first of its kind and presents quantifiable, albeit perhaps limited, data on correlation between the measured vibrations and failure rate (MTBF or MTBR).