Article 20: WHAT HAPPENS WHEN A PUMP NO LONGER OPERATES AT OPTIMUM CONDITIONS (Part 2)
In June of 2003, we started to discuss hydraulic implications of pumps operating to the left of the best efficiency point (BEP). Low efficiency, high radial loads, noise, vibration - become a real problem when that happens. Damage to the seal, shaft, couplings and poor reliability are a real and direct result of such operation. However, is it possible to quantify “Reliability”? What is the impact on the equipment Life Cycle Cost, when a pump operates, say, 40% off-peak? How much does this cost the plant? – not just in energy alone (that aspect we covered in June), but – in terms of … seal replacement … bearing life … coupling repairs … cavitation and recirculation damage cost … and so on. Is it possible to derive at some factors which would relate the maintenance dollars spent – to the inefficiency of the pump operation?!
In Part 1 (Article 19), we talked about the effects of pumps operating at off-peak flow on efficiency and did an estimate of the wasted energy. That discussion now has been published as so please feel free to examine it there. This month, we are now examining the effect of such off-peak operation on radial load, cavitation damage, and other aspects – and linking these to the estimate of the actual plant costs. To do that, some assumptions had to me made, and we would like to hear from you if you agree, or disagree, with these assumption. As times goes on, this fledgling theory of operation-to-reliability costing method may develop into a more comprehensive method, and the input from the users will help make the next step.
In the recent years, the importance of improving the overall reliability and plants uptime resulted in renewed attention to the reliability assessment of the individual system components. Pumps constitute one of the major classes of the plant components, and directly contribute to the overall economics of the life cycle evaluation.
Let’s take a look, for example, at the advantages of the use new materials, such as structural composites, with regard to parts upgrade program (pump impellers, wear rings and bushings), as well as complete pumps. The reliability and savings can be achieved with focus on the following:
Significantly reduced weight (80% lighter then metal) with excellent tensile strength, approaching metals
Superior chemical resistance for most demanding tough chemicals
Abrasion resistance – up to 15% solids
Dry-running – continuous 3-D interwoven woven graphite fibers providing good self-lubricating properties
Cavitation resistance exceeding bronze and stainless steel
Improved rotordynamics, longer seal life and bearing life extension
Quality: these thermoset parts 5-axis machined from solid block (no unbalance), as compared to injection-molded or cast thermoplastics (voids and crevices result in unbalance)
Higher efficiencies due to tighter clearances allowed and superior finish
Let’s also compare open versus closed impellers. Open impellers have traditionally been accepted as a standard design configuration of the ANSI pumps for chemicals. Initiated as far back as 1961 (originally called an AVS Standard), ANSI pumps have been installed in numerous plants, and established a convenient and accepted standard to which both manufacturers, and the end users, could readily comply. ANSI pumps, made by different pump manufacturers would adapt to the same piping and baseplate dimensions, thus making them dimensionally interchangeable, although the internal parts geometry differs from one manufacturer to another.
The main advantage of the open impeller design is cost. They are easier to cast and clean up at the foundry, especially in case of sand casting process, which is often applied to iron construction. Stainless steel designs are typically made using precision patterns, and cleanup of passages from the mold residue is less of an issue, but still a less expensive operation.
With the advent of the mag-drive designs, however, closed impellers became a necessity, mainly due to a need to reduce the axial thrust. Carbon or silicon carbide thrust washers, lubricated by the products, do not have the same load capability as antifriction ball bearings. Thus, a need to reduce the loads necessitated the change toward the closed impeller designs.
However, substantial reliability benefits can be achieved by replacing open impellers by closed impellers, - even for the standard designs that utilize single or double seals, or packings. Although the benefits of this approach have always been understood, the quantifiable justification has become possible only recently, as new reliability methods and approaches became available, using a Life Cycle Cost Method (LCM).
The benefits of the closed impellers are materialized in (4) following areas:
As a typical example, consider a typical MTX frame ANSI end suction overhung open impeller pump. Typical radial load FR = 400 lbs and axial load is FA = 900 lbs. The equivalent dynamic load is calculated as:
P = XFR + YFA = 0.63 x 400 + 1.24 x 900 = 1368 lbs
This load is carried by the 5306A double row ball bearing, which has a dynamic load coefficient C = 16,400 lbs, which results in L10 life calculation as:
L10 = (C/P)3 x 106/(60xRPM), and adjusted by the a23 coefficient, reflecting oil optimization (typically a23 = 2.5)
Thus, Lna = 2.5 x (16,400/1368)3 x 106/(60x3600) = 20,052 hrs, i.e. 2.3 years.
ANSI spec requires 17,500 hrs bearings life, which is approximately in agreement with calculations.
However, the axial load (FA) used in these calculations, can change dramatically, depending on the position of the impeller within the volute, the height of the pump-out vanes (POV), and the gap between the pump-out vanes and the casing wall. The design value, used at above calculations, is at the assumed design value of 0.060” gap between the POV and the wall. The main reason to use pump-out vanes (POV) is to change the pump axial hydraulic thrust:
The rotation of the impeller results in “dragging into rotation” of the fluid in the gap between the impeller and casing walls. This is similar to a motion of a teaspoon in a cup, or a disk spinning inside containment. The resulting motion is referred to as “forced vortex”. Such vortex sets-in in the front and back gaps – between the casing walls and the impeller front (shown on the right) and a back hub (shown on the left). The pressure distribution in the gap is parabolic – higher at the impeller OD, and gradually reducing towards the shaft centerline.
Pressure time area is force – which is exerted on the impeller from both sides (FR and FL). The difference between these forces is hydraulic axial thrust, which is ultimately transmitted to the bearings, and thus is desirable to be small. This pressure at a given position in a gap depends on the radius, rotational speed of the fluid (divided by the impeller rotational speed), and the gap.
Curve (1) shows the static pressure distribution behind the impeller hub without the pump-out vanes. As we know from basic hydraulics, - the faster the fluid moves, the lower the static pressure is. So, if we could make the fluid in the gap to rotate faster, the static pressure would be reduced, and the force FL to become smaller – closer, and hopeful equal to the FR – to reduce, or eliminate the net thrust.
Without the POV, the fluid in the gap is rotated only by the friction (drag) of the impeller hub wall. It rotates with the same speed as the impeller right at the impeller wall surface, but (so called “no-slip-condition”) is not rotating at the casing wall, since that wall is stationary. Thus, on the average, the bulk of the fluid in the gap is spinning at the angular velocity equal to half of the angular velocity of the impeller.
But, if we add the POV, the fluid becomes “trapped” within the POV space, and thus rotates with the same speed as the impeller – i.e. double of what the fluid does in the absence of the POV. This, of course, assuming the gap between the POV and the casing wall is (theoretically) zero (x=0). The number of the POVs actually does not have to be equal to the number of the impeller main blades, but often is, due to casting production process, for simplicity.
Obviously, the gap “x” can not be zero, so the actual reduction of the pressure profile (curve marked as “2”, is less, depending on the gap “x”. And, if this gap becomes too large, the effect of the POV diminishes, and eventually disappears. It turns out that, the POV are most effective at x=0, and become completely non-effective when x=t, i.e. when the liquid gap (x) becomes equal to the height of the pump-out vanes (t). (Papers are written on this subject, such as a well known Zanker’s paper), explaining whys and whats, and those of you who have a couple of sleepless nights to study them – let us know, and we will get you the material).
The balancing holes are also used to reduce pressure distribution, i.e. a similar idea as the pump-out vanes. In fact, this is why they are called “balancing”. To be effective, the impeller must have a tight clearance between it and a casing (not shown on a picture above), to separate the higher pressure zone, from a lower pressure zone. The balancing holes thus connect the back of the impeller with the inlet area, where pressure is low (close to suction). Some leakage will occur, reducing the efficiency. If there is no clearance, such as shown above, the leakage will be greater, reducing the efficiency even more.
Another reason for the POV is to reduce the pressure at the mechanical seal area. The effect of these vanes can be very strong, and, sometimes even result in creating vacuum, and boiling out of the liquid. This could be trouble, as mechanical seals do not like to operate in a vapor environment.
Regarding the performance – there is price to pay for the thrust balance, as the additional power required to spin the liquid faster takes away the efficiency. This is why, the higher energy pumps, such as API, or boiler feed, rarely have the pump-out vanes, while the ANSI pumps, which are relatively lower horsepower units, have these.
Note the picture above shows an open impeller. The front gap, between the impeller and the casing, must be tight (typically 0.005 – 0.015”, depending on a pump size). The challenge is thus to maintain both front and back gaps small – from the efficiency and thrust standpoint. Closed impellers solve this problem, but at the expense of reduced ability to handle stringy, such as fibrous, solids.
Regarding the “reverse vane” impeller. The impeller shown above is a standard design, such as, for example, manufactured by Goulds, model 3196. To adjust the front clearance, the rotor must be pushed against the casing, and then backed-off by the amount of design clearance. It is sometimes desirable to keep the casing piped-up, as it is somewhat a chore to re-pipe it. This requires re-setting of the front gap, during maintenance, on site – and, if it rains or snows – you freeze and catch a cold!
the impeller is “turned around”, such that the clearance gap is between the
impeller and the stuffing box (or a sealing chamber), then this clearance can
be set at the shop, and the rotor can then be simply brought to a casing and
bolted on quickly. Example of such design is former Durco
Clearly, the perceived advantage, widely publicized in the past by the manufacturers, of the ability to adjust the impeller front clearance to compensate for wear is quickly, can do more harm then good: as impeller is adjusted forward to close the worn out gap, the back clearance increases – by the same amount. The increased gap between the POV and the casing wall can effect pressure distribution dramatically. The resultant axial thrust can actually triple, i.e. FA = 3 x 900 = 2700 lbs
Additionally, as impeller wears out, the removed metal causes radial unbalance, which adds to a hydraulic thrust. Estimates vary, but 50% increase in radial load is quite possible, so that FR = 1.5 x 400 = 600 lbs
The equivalent dynamic load then becomes: P = 0.63 x 600 + 1.24 x 2700 = 3736 lbs, which impacts bearings life dramatically:
Lna = 2.5 x (16,400 / 3736)3 x 106/(60x3600) = 979 hrs = 1.3 months !
The impeller wear, and the resulted unbalance, results in increased radial load and shaft deflection:
Seal manufacturers have done research which shows exponential deterioration of seal life, including leakage and failures, when the angular misalignment exceeds 0.002”. At 50% increased radial load, a normal projected seal life of 2 years can be reduced to less then 6 months.
The problem can in fact become worse if the impeller is adjusted (as was discussed above) axially due to wear, because this results in reduced seal spring tension, lowering closing force and initiating premature leak.
Theoretically, an open impeller with negligible front clearance is somewhat more efficient then a closed impeller. However, efficiency drops very quickly as actual front clearance is increased due to wear. When initial 0.005” front clearance is opened up to 0.010”, a pump can loose 10% efficiency due to increased leakage, and at 0.015” there would be approximately another 10% efficiency reduction. Obviously, none of this is an issue for closed impellers.
The impact on NPSHR is similar to the efficiency. The increased front leakage effects the NPSHR and could add another 2-4 feet when clearance doubles. The R-ratio (NPSHA/NPSHR) varies form one installation to another, but a typical rule is to have at least 5 feet NPSHA above the NPSHA required. When almost half of this margin is taken away to front leakage, the R-ratio can become too close to the NPSHR and the cavitation bubbles that begin to form even before the NPSHA reaches the value of performance loss, become active enough to cause caviation damage:
It is difficult to quantify the decrease in impeller useful life due to worsen cavitation condition, since there is no known statistics published on the average life of the impeller as function of NPSHA, and some reasonable assumptions may be required. If we assume that a non-cavitation impeller could last 10 years until replacement, the reduction of the NPSHA margin from 5 feet to 2.5 feet would affect the life in direct proportion, i.e. reducing it to 5 years. More studies, however, would be required to quantify this particular effect on reliability.
LIFE CYCLE APPROACH
In order to compare different designs using Reliability methods, a Cycle Basis needs to be established, which is also applied to other types of components. A reasonable basis for comparison is a 5 year interval. Another assumption is made that, during this time cycle, a pump operates in several descretised steps of operation, for example 20% in regime 1, another 20% in regime 2, and so on. This allows a more uniform distribution of various regimes of operation throughout the cycle and a better averaged comparison. We will apply this method to each of the selected elements that are affected by the forcing function, such as:
a) Bearings – life decreasing from 2.3 years to 1.3 (0.21 years) months within five zones for the purpose of averaging:
The above distribution can be a safe assumption when a continuous statistical data is limited. Otherwise, if the load is known at each interval, and the bearing life can be calculated at each of the intervals, the averaging can be done by summing the life at the intervals and dividing by the total time span (5 years in case if Cycle Basis is selected as 5 years).
For linear distribution, the averaged weighted life would be:
(2.3 + 1.75 + 1.2 + 0.65 + 0.1) / 5 = 1.2 years
b) Seals – using similar linear function (2 years life at interval 1, reducing linearly to a calculated 0.6 years life at interval 5), we get:
(2 + 1.62 + 1.25 + 0.88 + 0.5) / 5 = 1.1 years
c) Efficiency - assuming pump efficiency drops from 70% to 58% (70, 65, 60, 55, 50), we get the weighted average:
(70+65+60+55+50) / 5 = 60%
To relate this to the life cycle factor, we need to evaluate the energy savings due to efficiency difference. Assuming a pump with a 10 hp motor, running 100%, at $0.06 per kWxhr, this results in:
10 x 0.746 x 24 x 365 x 0.06 x 0.70 = $2755
10 x 0.746 x 24 x 365 x 0.06 x 0.60 = $2352
Annual energy savings $402 per pump, which is roughly a 1 year payback period basis a typical cost of a replacement impeller.
d) NPSH margin – assume reducing from 5 feet margin to 2.5 feet (5, 4.4, 3.8, 3.1, 2.5), we get the averaged cycle value of:
(5+4.4+3.8+3.1+2.5)/5 = 3.8 feet
If a 10 year life is assumed with a full specified margin (5 feet), then the reduced life is 3.8/5 x 10 = 7.6 years
Note: the actual life of impellers also varies depending on applications and pump types. Double suction cooling water pumps, for example, are notorious for NPSH-related problems (Ref. ).
Bearings life factor 1.2 yr 2.3 yr
Seals life factor 1.1 yr 2 yr
Efficiency factor 0 yr 1 yr
NPSH life factor 7.6 yr 10 yr
Total: 9.9 yr 15.3 yr (54% improvement)
Having established life factor comparison, the next step is relate these to the Mean Time between Failures (MTBF), which is typically known relatively accurately for a given plant. Assume for example, a plant averages 3 years MTBF. The 54% improvement, calculated above, would increase the MTBF to 3 x 1.54 = 4.6 years.
What does an improvement in 1.6 years, per pump, in MTBF mean to a typical plant?
1.6 added years is equivalent to an improvement of 1/1.6 = 0.62 failures pre year. The cost of equipment failures is a total of individual components: parts, labor and lost production. While replacing of impellers, seals, bearings, and other parts is costly, a cost of lost production can even more significant. For a typical chemical plant with thousands of pumps, this could be a very significant number. The cost of a lost production also varies from one plant to another, and literature has values anywhere from $5,000 per hour to $200,000 per hour. At a $10,000 per hour value of lost production, each pump would result in 10,000 x 0.62 = $6,200 difference, and, with a population of 1000 pumps at a plant, this amounts to 1000x6200 = $6.2M for a plant!
Clearly, these vary from one plant to another, but the importance of the issue is clearly there, and it can not be ignored.
Example above: metal impeller, reduced life due to corrosion attack. Composite (continuous graphite fibers in epoxy matrix), not affected by chemical attack.
Also, an ANSI pump, installed at the chemical plant, resulted in extended life due to lowered weight, as well as improved chemical resistance. These cases are currently being prepared, and will be included with the final manuscript.
A retrofit program of converting open impellers to closed impeller design present substantial, real and immediate benefits to the end user. For a typical chemical plant, or similar operating facility, this could be thousands, or even millions of dollars, saved in maintenance, repair and production budgets. The conversion process is straightforward and technically sound. The best approach is to establish a planned program of replacing, starting from the worst operating units and continuing to the next level, driving the plant reliability record continuously upward.
New technologies, such as engineered structural composites, present a timely and effective opportunity to achieve maximum benefits and quickly. The combined benefits of significantly reduced weight, substantial improvements in chemical resistance, excellent abrasion characteristics, and superior cavitation resistance, makes Simsite a material of choice for such program.
We welcome your input on this topic, and will be glad to post your comments and opinions at the Pump Magazine On-Line commentary board.
Dr. Lev Nelik, P.E., Apics
Pumping Machinery, LLC