People,

I have been reading the e-mail's with great interest as this has always been an issue of great importance to me. Standby equipment performing to specification when required is a very important issue in the petrochemical industry, when continuous processing and series lines of equipement.

I agree with the concept of having a standby and just putting it in operation for failure finding purposes, the 90:10 rule.

We have gone one step futher and used visual aids to mark the standby equipment with a bright colour. Every and any person walking through the plant will notice the odd colour equipment in operation and ask questions about when the plant condition will be restored to normal.

On isolated sites in Mining we also found this to be very important, as long lead times to repair equipment exposes the plant and reliable standby equipment become very important in this context.

Regards,

Neil Blom
Reliability practitioner

Nice idea about the color. At least if takes out the guess work which equipment is duty or standby.
Three pumps on a row, when one can do the work, will result in canibalism.

I see...
First Pump: Main one, Duty taker, this one canibalize the other two.
Second Pump: Stand by, canibalize just from the third.
Third Pump: Just a storage location placed at the side of the other two, only purpose in life: keep parts handy for the first one and the second one, hours of operation near to zero.

I think cannibalism will happen because the spare part replenishment process cannot keep up with the maintenance requirement. I guess it' necessary to support the duty-standby policy with efficient spare part acquisition process.

I still don't understand why there is an assumption that 50/50 runtime gives lower reliability than uneven splits (90/10 etc).

Please explain.

Hi electricpete,

Normally the 50/50 runtime strategy causes both pumps to wear at the same rate. In a well designed and managed system this will cause the 2 pumps to fail at the same time. With a site with long lead times to deliver the spare or all the spares being out for repairs, it may cause production losses. It is not a reliability issue, it is an issue of how the reliability is managed.

What about using a 90/10 strategy for a year and after the annual shutdown switch the pumps and start with the pump B running 90% and pump A running 10%?

Pete, Darth,

The worst condition for a pump is when you start it. Seal failures, which are strongly affected by the number of starts are less likely when you keep the pump running.

We are interested in system availability; i.e. at least one of the two pumps should work well. By keeping the running hours on the standby as low as possible, we ensure that it sees very little wear and stays in pristine condition. When you really need it to work, you know it is nearly as good as new. This way you can be more sure that it will work.

But an idle standby pump ahs an important(hidden) failure mode. It may not start, or if it does, it may not take the full load. Hence we have to start it, put it on load for a few hours to satisfy ourselves that it will do its job when required. We can do this through a test start or by a 90:10 policy. Darth, flipping the 90:10 defeats the objective of keeping one pump in pristine condition.

It can be shown that with a duty standby policy,
- the total number of failure modes to address for the two pumps is less than with other policies
- the system availability is higher than with other policies

90:10 is close to a duty-standby situation.

I hope this explanation helps. If you have a copy of my book, see a more detailed explanation on page 168.

"It can be shown that with a duty standby policy..."

How is it shown? What is the nature of the proof and underlying assumptions?

Once again some mention of starting duty. Starting duty is the same for either mode 50/50 or 10/90 (if not worse for the 10/90 for reasons I outlined in my previous post).

There was mention of machines getting less reliable toward the end of their refurbishment period. Is there an assumed refurbishment of the standby on a time-basis? On a run-time basis?
We have ability to completely control the overhaul schedule. If concern is to minimize the number of refurbishments without getting two motors near end-of-life, it seems this situation is much more manageable in the 50/50 scenario if the refurbishment intervals are staggerred among the two pumps. With 90/10 interval I have no idea how the refurbishment intervals are planned but it doesn't sound any easier to me.

As far as the idea of a "pristine" standby machine with almost no run-time, that doesn't sound particularly attractive from the standpoint of identifying infant failures at the right time (before you really need that motor to start in standby).

Pete,

Please read the section on duty-standby on pages 168-170 of my book. See also John Moubray's book RCM2, pages 28-29 and 105-110. If there are doubts after that, let us resume the discussion.

By the way, this is not just theory, I have app;ied it and measured the difference! The operting philosophy has a major effect on system availability and maintenance costs, see article in the main Reliabilitycom website on 'Opertaing Philosophy'. Regards.

Vee from a practical point of view, a new plant (new operators, equipment) running 24 x 7.

In the first 3 years we had the philosophy:

Run the main pump, mostly pumps with hot products (it is a refinery), and once a week, run the standby pump one shift (8 hours).

We had "abnormally" frequent mechanical seal failures. After abolishing this policy, and only switch to the standby pump when performing PM, and when absolutely necessary the failure rate went down, at least that is our impression.
Please comment how did you conducted your experiments. I am in favour of the 90-10 approach.
I must also admit that in our inexperience we did a lot of "wrong" things.
By the way, I have your book, my only remark is that the Alpha Piper disaster is very well documented and I wished the Bhopal incident had the same treatment.

Regards

Vee I did a search on operating philosophy (yes I spelled it correctly), any links?

Svanels,

quote:

We had "abnormally" frequent mechanical seal failures. After abolishing this policy, and only switch to the standby pump when performing PM, and when absolutely necessary the failure rate went down, at least that is our impression.
Please comment how did you conducted your experiments. I am in favour of the 90-10 approach.

Did you get to the bottom of why the hot pump seals failed? I suspect but do not have data (but you may have it) to support my hypothesis, which is
1. Weekly starts mean 52 starts of each of the duty and the standby pumps, i.e 104 for the pair.
2. Every start sees severe operating conditions, with the highest seal wear.
3. Most Refinery pumps, especially in hot services need a lot of care during startup; seal flush, cooling water to stuffing box or seal fluid cooler, venting of stuffing box, steam quench etc. It is easy to get one or two things wrong on some occasions. This will start rapid degradation in those units. Overall, it appears that all pumps suffer, but they may be due to a variety of reasons, as discussed here.
3. I am guessing that your PM checks were monthly. With starts of standby during PM checks of duty pump, you had 12 starts of each pump, or 24 for the pair. That is a lot less than 104! 80 more starts means 80 severe loading periods plus 80 opportunities to have operator errors.

You also stated, quote
Run the main pump, mostly pumps with hot products (it is a refinery), and once a week, run the standby pump one shift (8 hours). unquote

To minimize 'fail-to start' incidents, in most oil services, for horizontal pumps, a 2-monthly start will give you a reasonable level of availability (97% or more, depending on actual failure rate). With this policy, each pump will start 6 times, or 12 for the pair. With a 90:10 type policy, say 8 weeks on 1 week off (88:12), you get only about 8 starts in all for the pair. So you can see how the operating policy affects the number of starts with all its problems. I am of the view that failures per no. of starts is a better PI than failures/operating hour. That is another subject!

Your policy of keeping the standby on for 8 hours was quite right, it is just the weekly starts, I think, that killed the pumps.

I too did my trials in a Refinery, in what I believe is one of the most aggressive conditions; Bolier Feed Water Pumps. I will be describing this in a new book I am writing now. After implementing a 75:25 policy on all pumps, many of which were in very hot oil services in e.g.,(Hydrocracker, Thermal Cracker services, our failure rates and costs dropped by about 7-8% while the availability rose by about 10%.

If you adopt a 90:10 policy, you should see a marked change within 3-4 months, if the pumps are in relatively good condition now.

The link is http://www.reliabilityweb.com/art04/operating_philosophy.htm

I too had difficulty locating the link; finally used Google with the words operating philosophy reliability.

Regards,

Vee, we have 2-monthly PM-checks on the pumps, it was a battle to convince operations to abolish the weekly 8-hours run.
Finally we convinced them with the theory, that the pump wasn't loaded to its full capacity in the short hours, from hot-standby to full service.
We could track a paralell to a generator that was working only 60% of its rated capacity and was giving more headache than work.

I am looking forward to your new book

How often should you test run the spare pump? Weekly sounds far too often.

Apply the RCM philosophy. The function to be preserved being to be able to take over the function of the duty pump in the event of duty pump failure.

One major global producer appplied this to their pumps and determined that the frequency required to test run most pumps is once every three years (for an availability target of 97%). The driver for this being seal binding due to prolonged period on standby (MTBF = 50 years).

Try the same exercise for your pumps and see what period you come up with.

Vee you made my day!!!

We are in the stage of implementing a vibration monitoring program.

We (maintenance) were discussing how are going to take the vibrations on the standby pump, if we had only lets say 2-4 hours to do the pm-check

We do also a 6-month alignment check.

The 1 week duty would give us time to evaluate the standby pump on normal working conditions, in a steady state.

Hm.. another battle, to convince the "owner" that he has to run his standby pump for a week. In other words "dictating how they must run their plant"

By the way,the 8 weeks on - 1 week off gives 90 - 10% (6 starts for the main, 5 for the standby)


The availability of 97% I have difficulty to see the math, Probability and Statistics was not my favourite subject because the examples we had in class were more about what the probability was to get a six when throwing 2 dices.
And since I am not the gambling type...

Probably I have to work out your example in the book..

This message has been edited. Last edited by: svanels,

If I had to test run the standby pump once in three years, it probably would be canibalized in the first year or the bearings would have suffered from false brinelling because of the surrounding vibrations

David,

quote:

One major global producer appplied this to their pumps and determined that the frequency required to test run most pumps is once every three years (for an availability target of 97%). The driver for this being seal binding due to prolonged period on standby (MTBF = 50 years).

I do not understand the 50 year MTBF. In my experience, pump MTBFs (mode:stopped while running) range from about 12-60 months.

As far as the standby is concerned, the MTBF we want is that of 'fail-to-start'. Usually this is measured as no.of failed starts divided by total number of starts. It is a %age figure, that gives the start unreliability directly {or F(t) from which we can get the R(t) as 1-F(t)}. If we are interested, we can compute the start MTBF from this value, assuming an exponential distribution. I dont think you will get 50 years from this exercise. Using data from generic sources like OREDA, test starts should be at 1-3 months, not more. Perhaps you can explain the source of the global producer's data. By the way, seal binding is not the only mode to consider, there are other reasons for 'fail-to start' too.

Svanels,

quote:

Hm.. another battle, to convince the "owner" that he has to run his standby pump for a week. In other words "dictating how they must run their plant"

I see you have a them-and-us problem. Surely it is 'our' plant so we only need to clarify how the duty-standby policy will help 'us' get higher availability at lower costs.

Running the standby for a week gives you time to take condition monitoring readings on the standby units. Without this policy, it is hard to get regular readings, it is always hit or miss. There is another benefit, as explained below.

When we first introduced the 75:25 policy (let us say the 'A' units were on the 75%), I noticed during morning rounds that some 'B' units were running. Checking with the Control room operator, I found out that the relevant 'A' units had some problem in the earlier shift(s), BUT there was NO breakdown work order YET! Aha, why so? Oh, we will get around to it, he said. That identified one thing; operators dont always REPORT failures AS SOON AS they occur. With the earlier policy, they may not report it for several days. This means we lose response time to plan and execute. Now with this new policy, they could not hide these failures from maintenance! Transparency helps performance improvement. With quicker reporting, maintenance responded better, and uptime improved. Operators could see this clearly, and this helped reduce the them-and us situation. It helped that i got on very well with the Operations mnager and his deputies.

I wish you luck in breaking the Silos.

quote:

Originally posted by Vee:
Please read the section on duty-standby on pages 168-170 of my book. See also John Moubray's book RCM2, pages 28-29 and 105-110. If there are doubts after that, let us resume the discussion.

Sorry I don't have those references. Can anyone provide a brief discussion on what basis there is a belief that 90/10 is better than 50/50. I have not yet seen a shred of answer in this thread.
Thx