MTBF for "centrifugal" pumps

[Ayden]

I'm a Reliabiilty Engineer and working now in the refinery for 6 months. My question is how do Refineries calculate MTBF on their pumps? Do we need to count all installed pumps in equation or only Running pumps? For example in a Turnaround should I still count the pumps that are down in that paticular month? Any answer to my question would be appreciated .....

 

[tonyh]

A very good question to which there is no standard answer. It all depends, and it demonstrates why comparing MTBF between different organisations is fraught with trouble.
It depends what you wish to achieve. My own view is that it is does not matter much provided that you are consistent. MTBF is best used for seeing whether you are improving or whether you are getting worse. In that case the base number of pumps is not a big issue unless you change part way through. If you wish to compare yourself with another user, it is good to use their definition.
On our chemical site we use all installed pumps on the grounds that when they are installed they are subject to wear and tear from vibration, washing down, etc. Also we do not have a great number of installed spares.
There are a lot of other variations in definition for MTBF around what constitutes a failure. We get blockages and minor maintenance issues (we call interventions). If we clear a blockage and don't overhaul the pump we do not count this. We overhaul some pumps routinely but most predictively. We count all overhauls of these sorts as "failures" for our monitoring on the grounds that the predictive pull is for an incipient failure and if there were no degradation mechanism we would not be doing routine overhaul. Other people have different conventions.
Another issue is that there is a variation in standard of what people will accept as a running machine. Does a minor barrier seal fluid leak cause you to pull a pump? Depending on your standards you get different unit lives.

 

[smckennz]

Because a pump that is not running cannot really "fail", it would appear to make more sense to base your MTBF on running pumps. For example if you have only one pump and it fails daily, the MTBF is 1 day. If you install another 9 pumps which do nothing and never run, the MTBF increases to 10 days, on a total pump count basis, although there is no change in plant reliability.
The exception would be if it was possible for non-running pumps to fail e.g. in a highly corrosive environment.

 

[kris44]

I agree with tonyh that MTBF is a measure of if you are improving or not. The best way to get a MTBF is to group pumps with similar characteristics (same fluid, similar operating times, etc) and then calculate a MTBF for each group. If you go back through your records, then you can see if the practises that you're putting in place are working or not.

Also you may find in some (particularly in corrosive) environments, that the less a pump runs, the lower the MTBF.

 

[JJPellin]

I have been responsible for tracking the pump MTBF in a medium sized oil refinery for most of the past 15 years. There are several good points in the replies above. We count all installed pumps whether they run or not. The only exception would be if the pump were removed from service (blinded or disconnected). Our in-place spares are not running but are usually flooded with product and are started once per month to be sure that they are ready to run. We definately do have failures in these pump even though they don't normally run. The tricky part of calculating MTBF is the definition of a failure. We coun't any work by a mechanic on a pump as a failure unless it is a preventive or predictive maintenance task. Even if the vibration program identified a bearing defect long before the bearing was noticeably vibrating, we still count it as a failure. It was in the process of failing. We just caught it early. Some plants only count a failure if they had to replace parts. I even heard of one plant that didn't count failures unless the pump had to be brought into the shop. The mechanics learned of this definition. They mounted vices on the backs of their trucks and would perform full pump overhauls out in the field to avoid having them counted as failures. The most important thing is to set a definition and use it consistently. As noted above, you are only comparing your reliabilty to past reliability within your plant. I never trust numbers from other plants to compare to my own. There are too many ways to fudge the numbers. Good luck.

 

[25362]

Everything has been said concerning ways to estimate MTBF. As to how to improve the MTBF for centrifugal pumps, including all their component parts, the consensus is that the best option is to operate as near as possible to their BEP. Chances are that levels of 20,000 to 30,000 hours of continued operation could thus be achieved.

If that is not feasible, there are other recommended methods to help out in prolonging the MTBF, among them:

-reduce impeller diameter;
-install by-passes;
-have speed-controlled drivers;
-control flow rates with automatically operated valves;
-put holes thru impellers to equalize pressures.