Centrifugal Pump Impeller Failure

[JJPellin]

I have seen a problem with two of our centrifugal pumps and am curious if anyone has seen this before or has an explanation for possible cause. First, I should provide some information about the pumps and the service. These are single-stage, overhung, centrifugal pumps that operate at 3600 rpm. They have enclosed impellers with three curved vanes. They have wear rings on the head and case side and three large balance holes. They are very typical API pumps built in about 1999 by a major manufacturer. The impellers are 14.125 inch diameter. The maximum impeller diameter for this model is 15.625 inch. The impellers, heads and cases are all cast 316L Stainless Steel. The shrouds on these impellers are about 3/16 to 1/4 inch thick.

The pumps are in lean amine service at about 140 °F. They are typically operated at about 1000 gpm and 820 feet of head. The product specific gravity is 0.98 at normal operating temperature. The suction specific speed is just over 13,000 in US units.

The problem was see on the bench today has occurred three times in 10 years of operation. The shrouds on the impeller are bulged out between the vanes to the point where cracks are developing in the corner where the vane meets the shroud. The first time this occurred, it was on the spare pump that rarely runs. The impeller was replaced in kind since none of us had ever seen this before and we had no explanation for the cause. The second time also occurred on the spare pump about 4 years later. This time, the damage had progressed to catastrophic failure on the head side. At this failure, we replaced the impeller with one we had designed to incorporate short half-vanes at the OD to support the shrouds half way between the full-vanes.

The third failure occurred on the main pump that normally runs 24/7. With the current failure, the shrouds are bulged out about 1/8” and cracks are just visible at the corner where each vane meets the case-side shroud. The bulging looks like it is about the same on both shrouds. The impeller in our shop right now also has significant cavitation damage in the eye on the visible side of the vane back about 3/4” from the leading edge. I don’t believe we saw any cavitation damage on the previous two failures.

I am interested to know if anyone else has seen failures of this type. I would also, very much like to hear your opinions as to possible cause.


Johnny Pellin

 

[MikeHalloran]

With only three vanes at that diameter, the shrouds are equivalent to moderately thin quadrilateral plates supported on two sides, possibly stiffened on one side. It might be interesting to estimate how much internal pressure it would take to reach the yield stress at the supported edges, probably around 27ksi. I.e., I'm guessing maybe a pressure excursion. ... but that should produce bending, without cracking.

Or maybe the cracks came first, e.g. from resonance of the shrouds, and eventually converted the edges from fixed to pinned, allowing the deflection.

I've never seen it, but I'm guessing your added half-vanes should fix it. We should know in a few years...




Mike Halloran
Pembroke Pines, FL, USA

 

[tr1ntx]

"The first time this occurred, it was on the spare pump that rarely runs."
Do you have any idea how much that pump had been run up to that point?


"The second time also occurred on the spare pump about 4 years later."
How much had it run this time? Were the failures noticed at the first maintenance cycle, or was the catastrophic failure the reason for opening it?


"The third failure occurred on the main pump that normally runs 24/7."
How long since it had last been checked?


The first 2 failures had me thinking probably bad castings (or possibly poor design of a cast part). This last one, hmmm . . . . It's possible cavitation-induced vibrations could cause the cracks. The amplitude would be greatest out at the OD and the stress would peak at the corner where the cracks are. But that doesn't explain why rarely used backups would have them.

How does the spare pump operate? Is it just brought online when one of the others is off, and has the exact same operating conditions, or is it triggered on by some event/parameter? The cause might be a pressure excursion like MikeHalloran said, a "water hammer" effect or something.

 

[JJPellin]

I have a little bit more information and answers to some of your questions. The spare pump is turbine driven and set up to auto-start, so they don’t like to run it. It runs at least four hours per month so that they can take vibration data under steady state conditions. Otherwise, it only runs if the main pump is out for repair.

The main pump runs continuously except for the 4 hours per month that they are running the spare. The main pump had been in operation for three years since it was last pulled. At that time, no impeller damage was noticed. This does not guarantee that there was no damage at that time. This is the sort of thing that our mechanics might overlook.

I am not sure of the history of the spare pump prior to the first failure. That failure was only discovered when the pump came out for a seal leak. What they found with the impeller was totally unexpected.

The catastrophic failure of the spare pump was written up as high vibration and poor performance. When they pulled the pump, they found the majority of the back shroud had broken up and was ground into a pile of scrap metal.

Our outsourcing coordinator spoke to an impeller fabricator we work with in Houston. He claims to see this problem regularly. They are the shop that recommended the half-vane option that we installed in the spare pump. They believe it is just pressure over area. This seems strange to me. The inside face of the shroud should see pressure starting near suction at the ID and increasing toward the OD. The outside face of the shroud should see the higher pressure all they way down to the wear ring. In other word, if I don’t allow for differences in velocity, the pressure should be greater on the outside and the shrouds should tend to collapse inward, not bulge outward.

Of course, I am not impeller design expert. I am not sure of how the pressure develops as the fluid moves out through the impeller. How much of the pressure develops in the impeller and how much develops after the flow enters the volute? But, anyway I consider it, I would still expect the pressure to be greater on the outside than the inside.

The Houston fabricator claims to see this on three vane impellers with relative large OD made in softer materials. This could explain why I have never seen it before. We have about 1400 centrifugal pumps. But, if I was looking for impellers with three vanes, larger than 12 inch diameter and made of 316 SS or something softer, there are probably only a few. Most of our impellers have five vanes and the most common material is CA6NM (12% Chrome Steel). After that, we have a lot of cast carbon steel and cast iron impellers. 316 SS is rare.

I have considered the shroud resonating and cracking from fatigue. But, the vanes have all been bulged outward with the cracks staring at the inside. The bulge is great enough that the vane is plastically deformed convex. This doesn’t seem to match that theory. And, the fact that the failures have happened more often on the pump with the least run time suggests it is not time-based, but pressure based. The pressure distribution on the two sides of the shrouds is probably much more complex than I imagine. I don’t have good information about the flow from this pump over time. If there are flow excursions high or low, these could result in different pressure distributions that are capable of deforming the shrouds.

Thank you for your replies. I will continue to dig and try to get some better information about the flow history of these pumps.

Johnny Pellin

 

[rmw]

Ahah, turbine driven huh?....

What is the operating speed of the turbine driven pumps? Is it also 3600 right on the money? And what is the overspeed trip setting?

I think you may see where I am going with this. What is the possibility that these pumps spool up higher than the motor driven 3600 rpm pump until the controls react and bring them back - IF the controls bring them back?

rmw

 

[electricpete]

Weird. Got any pictures?

=====================================
(2B)+(2B)' ?

 

[electricpete]

I am not a big pump guy, but just to throw out an idea and you can tell me if it's hosed. Let's say for some reason (*)there is a very very high recirculation of fluid which passes from the discharge, across the wear-rings, rushing radially inward across the outer sides of the shroud, and exiting either into the suction or some shaft seal area. Then the high velocity of that fluid flowing through the small clearances on the outer sides of the shroud creates reduced pressure both based on: 1 - Bernoulli/venturi effect; and 2 - friction losses. The lower pressure on the outside of the shroud creates the dp for bulding

* Now, why does the water flow in that pattern? Beats me. You guys might have some better ideas than me.

=====================================
(2B)+(2B)' ?

 

[JJPellin]

rmw,

Interesting idea, but I think I can rule out speed (rpm). When the turbine is run monthly for vibration analysis, the tech would verify that the rpm was correct (3600 rpm). The turbine trip speed is about 4050 rpm and is tested annually (three trips, non-trending, within +/- 2%). The trip test is performed uncoupled so the pump should never see that speed. During start-up, the speed may overshoot slightly, but it cannot reach 4050 rpm, or it would trip on overspeed.

Pete,

Velocity down the outside of the shrouds may very well contribute. But, the flow passing that direction does not pass a close clearance until it reaches the wear rings at the ID. The axial clearance between the head and the impeller is about 0.375". On the case side it is ever greater: probably 0.5". The wear rings have a clearance of about 0.008" to 0.012" radially. But, you may be onto a contributor. The fluid within the impeller must rotate with the impeller because of the vanes. Once the fluid exits the impeller, its rotation may be reduced, resulting in higher relative velocity over the surface of the shroud. But, even if I could imagine that the pressure on the outside of the shroud could be less than the pressure on the inside (which is a stretch), I cannot imagine that this pressure difference is great enough to deform the shroud in the manner I have seen.

I will try to get a picture that does it justice.


Johnny Pellin

 

[tr1ntx]

So it's possible the cracks developed in every impeller (original design) but only the one came apart.

I made a little sketch of a 14.125" dia 3 vane impeller, and good grief, that's a lot of distance between supports at the OD. I don't know the angle of incidence that the vane approaches the OD circle and how much "wrap" there is, but I guess it doesn't matter, there would still be the same distance. With the spare only running 4 hrs/month and having as bad or worse failure than the 24/7 main pump, that makes me think either bad castings or the fact that the design is inadequate to begin with. From what the fabricator in Houston said, it points to inadequate design.

I wonder if that impeller design actually has a lower speed rating, but that fact has fallen through the cracks with the manufacturer. I've seen that sort of "information failure" repeatedly over the years as companies have been bought up by these industrial conglomerates and implemented business-school-taught cost cutting, getting rid of the older employees who actually know this kind of stuff. Usually then consolidating product lines for the sake of operations streamlining. Then they probably off-shored their casting supplier, so who knows what you're getting.

 

[ash9144]

Could it be heat transfer down the shaft possibly boiling the liquid inside the pump leading to high pressure inside the impeller?

 

[JJPellin]

tr1ntx,

I think you may be onto a good point. I picture an impeller design that worked well for smaller diameters but has problems when it was scaled up. Or perhaps a design that was sound and worked well for cast iron and chrome steel, but was applied to SS without consideration for the change in material properties. I can picture the same thing for a design that worked very well at 1800 rpm, but was applied to 3600 rpm without the additional analysis needed. I might term this as a cut-and-paste design that was stretched beyond the original engineering and blindly over applied.

I am feeling more and more that this is just a basic design problem. As a point of curiosity, I want to understand the mechanism. But, when it comes down to brass tacks, perhaps I don't need to fully understand the underlying principles as long as I understand and can implement the solution.

I tend to rule out a simple material or manufacturing defect. I have had three impellers (at least) fail in this manner. They were made as much as 10 years apart, probably at different foundries and possibly in different countries. (The OEM moved most of their casting from the US to Mexico a few years ago) The only things I know for sure that they have in common are the design and the material of construction.

I am still very curious if anyone else has seen this exact phenomenon. With millions of pumps out there, I find it hard to believe that I have a problem that no one else has seen.

Thanks you all for the comments and discussion.


Johnny Pellin

 

[rmw]

Johnny,

Can you post any pictures? I'd like to see where the bulging is occurring. How large is the eye on the enclosed end? How far down toward the shaft do the vanes go?

I'm not sure I ever recall seeing a pump impeller that size with only 3 vanes and 3 vanes only (if that statement makes sense.)

In a former life I dealt with an eddy current brake rotor that was just a glorified pump impeller (vanes and shrouds to induce air flow for cooling) that had cracking problems at the root of a (6) vanes model due to the WR^2 of the impeller parts they supported. A similar wheel with more vanes with more 'windup' or 'spring' action in the vanes didn't have the problem in the same service.

rmw

 

[1gibson]

Shroud seems a bit thin at 3/16 - 1/4". Possible that cracks begin near areas that were ground on during balance operation?

3 vane vs 5+ offers better NPSH, less inlet obstruction. And based on Nss of 13,000, I'm sure your NPSHr is quite low compared to a similar 5+ vane impeller. So not very likely that a new impeller design would fit the application.

I'd go with inadequate design - thin shroud, insufficient material strength. Try upgrading from 316 to CA6NM.

If you have a single volute case, pressure spikes (high radial loads) could cause the bulging / fatigue of those shroud areas. This would be in the "3600 rpm oversight" category.

 

[JJPellin]

I have some additional information. First, what appeared to be cracks are not cracks. The second picture attached shows how these "cracks" appear at the tip of the vane. On close examination, these are erosion grooves cut into this location because the vanes are not straight, axially. They are angled, which creates a tendency for abrasive solids to follow a path along the underside of the vane right against the corner of the outer shroud. So, the impeller is deformed but not cracked.

As requested, the first attached picture provides a little better idea of the geometry, vane spacing, vane angle and wrap. Unfortunately, these are the only pictures I have and the impeller has been returned to service as an emergency spare while we wait for the delivery of a replacement impeller with additional support vanes.


Johnny Pellin

 

[JJPellin]

Sorry. One of the pictures was not attached to my last post.

Johnny Pellin

 

[d23]

JJPellin

I assume the amine is used to scrub H2S. Any possibility H2S is attacking the 316 SS?

D23

 

[JJPellin]

316 SS is the preferred material for all amine streams. The particular pumps I am discussing are in lean amine (very low H2S). But, the same material is recommended for rich amine (high H2S). All of our amine pumps are made entirely from 316 SS, including case, heads, impellers, shafts, seal parts, etc. I don't believe that H2S attack is related.

Johnny Pellin

 

[Artisi]

Johnny,
Questions - has the impeller bulged or is it "as cast", I wouldn't be that surprised to see this variation of impeller width if patterns were old or poor foundry practice or a combination of both. As for the crack maybe inbuilt casting stresses - the reason that nothing seems constant for failure rate / file.

 

[Artisi]

Johnny,
An after thought, do you have a new impeller to check measure for consistency?

 

[JJPellin]

Artisi,

They have definately bulged in service. The one we just had on the bench a few days ago was not extremely distortd. But, the last two were more extreme.

Johnny Pellin