Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations TugboatEng on being selected by the Eng-Tips community for having the most helpful posts in the forums last week. Way to Go!
API MTBF 1
- Thread starter jet1749
- Start date
-
1
- #2
This is a very difficult thing to compare between sites. We are currently running about 50 months between failures on our pumps. The population at our plant is probably 80 percent API. In the past our management has attempted to compare our MTBF to other refineries. These comparisons almost always make us look like we are doing worse than we are. By attending conferences like the Pump Users Symposium, I have found a great number of definitions of a failure (or repair).
It is very common to exclude adjustments that require no parts. So some plants only include workorders with materials costs. In that same category, there are plants that set a minimum dollar value for a workorder to count as a failure. In those, they might exclude any workorder with a total cost less than $500 or even $1000. The most extreme case I ever heard of was a plant that set up bonuses for their mechanics based on improvements in their MTBF. This plant only counted failures if the pump had to be brought into the shop for repair. If it could be fixed in the field, it was not counted. Since the mechanics had a bonus tied to the measure, they got creative. They mounted vices on the backs of their trucks and started doing full pump overhauls in the field so they didn't count.
Some plants count failures of the driver as part of the pump train. Some count turbine failures but not motors (since the electrical group is counting those). Some count problems with gearboxes, lube skids, mechanical seal support systems (seal coolers, pots, etc.) and some do not. Some include the piping up to the first block valve. Some exclude anything beyond the pump flange. Some plants wait until the Root Cause Failure Analysis is complete. If the RCFA indicates that the failure was caused by something the operators did (such as run the pump dry), they don’t count it. Some exclude bearing failures if the were found by vibration analysis.
We have taken an extremely conservative approach. If someone believe they have a problem with a pump, they enter a workorder, a mechanic goes out to the unit and gets a work permit and touches the pump, even for the most minor adjustment, it is a failure. Even if he did not replace a part, even if the problem was operational (low tank level) we still count it as a failure. If vibration detects a bearing that is failing, we still count it. We do not count driver failures (motor or turbine). We do not count piping problems outside the pump block valves. Overall, this is a harsh way to judge pump reliability. But it allows no room for fudging. The engineer who generates the scorecard has very little leeway to exclude workorders that have mechanic labor billed to them. But this way at least we are consistent.
So if others chime in with MTBF’s much greater than 50 months, I would be interested in hearing how they define a failure.
Determine a consistent definition of a failure. Determine your baseline reliability for the past 4 or 5 years. And then evaluate whether your reliability is going up or down. Getting a fair comparison to others (or any industry standard) is problematic, at best.
It is very common to exclude adjustments that require no parts. So some plants only include workorders with materials costs. In that same category, there are plants that set a minimum dollar value for a workorder to count as a failure. In those, they might exclude any workorder with a total cost less than $500 or even $1000. The most extreme case I ever heard of was a plant that set up bonuses for their mechanics based on improvements in their MTBF. This plant only counted failures if the pump had to be brought into the shop for repair. If it could be fixed in the field, it was not counted. Since the mechanics had a bonus tied to the measure, they got creative. They mounted vices on the backs of their trucks and started doing full pump overhauls in the field so they didn't count.
Some plants count failures of the driver as part of the pump train. Some count turbine failures but not motors (since the electrical group is counting those). Some count problems with gearboxes, lube skids, mechanical seal support systems (seal coolers, pots, etc.) and some do not. Some include the piping up to the first block valve. Some exclude anything beyond the pump flange. Some plants wait until the Root Cause Failure Analysis is complete. If the RCFA indicates that the failure was caused by something the operators did (such as run the pump dry), they don’t count it. Some exclude bearing failures if the were found by vibration analysis.
We have taken an extremely conservative approach. If someone believe they have a problem with a pump, they enter a workorder, a mechanic goes out to the unit and gets a work permit and touches the pump, even for the most minor adjustment, it is a failure. Even if he did not replace a part, even if the problem was operational (low tank level) we still count it as a failure. If vibration detects a bearing that is failing, we still count it. We do not count driver failures (motor or turbine). We do not count piping problems outside the pump block valves. Overall, this is a harsh way to judge pump reliability. But it allows no room for fudging. The engineer who generates the scorecard has very little leeway to exclude workorders that have mechanic labor billed to them. But this way at least we are consistent.
So if others chime in with MTBF’s much greater than 50 months, I would be interested in hearing how they define a failure.
Determine a consistent definition of a failure. Determine your baseline reliability for the past 4 or 5 years. And then evaluate whether your reliability is going up or down. Getting a fair comparison to others (or any industry standard) is problematic, at best.
Take back control of your plant from the bean counters. I don't think extreme is the right word.
1.) "and touches the pump, even for the most minor adjustment"
2.) "Even if he did not replace a part,
3.) "even if the problem was operational (low tank level) we still count it as a failure."
There is no way that I would believe that an operational problem or "touching" a pump such as,should be included in MTBE figures.
#1 is called normal maintenance.
#2 is called normal wear.
#3 is called insufficient operator training
BigInch![[worm]](/data/assets/smilies/worm.gif "[worm] [worm]")
-born in the trenches.
1.) "and touches the pump, even for the most minor adjustment"
2.) "Even if he did not replace a part,
3.) "even if the problem was operational (low tank level) we still count it as a failure."
There is no way that I would believe that an operational problem or "touching" a pump such as,should be included in MTBE figures.
#1 is called normal maintenance.
#2 is called normal wear.
#3 is called insufficient operator training
BigInch
-born in the trenches.MTBF...isn't the F for failure, the measure of which being when something doesn't perform to the degree expected. In the context of mechanical equipment, BigInch is consistent with the refinery culture I'm most familiar with. I'd like to say it was as simple as whether the equipment came out of service on its terms our terms, but there’s too much grey area.
I don’t think preventive maintenance programs or activities should count as failures but you need to be honest about it. Don’t move up a scheduled oil change due to increased vibration readings. With enough equipment history, a good preventive maintenance program should address issues that could causes failures, assuming quality material and craftsmanship are part of the program. Intentionally having unnecessarily frequent preventive maintenance activities can be cheating, too.
Predictive maintenance programs can easily trigger equipment downtimes that avert failures and, in my mind, should be counted as a failure since it a type of damage control. I’d gladly take a 2-day “failure” to correct something then a 2-week failure to replace something.
All these types of programs and measures have very noble intentions. Unfortunately how they are applied and the pressures imparted to get the numbers “right” can distort the entire process and cause a lot of headaches and misinformation. BigInch’s bean counter reference is right…they need to stick with cooking their own books.
I don’t think preventive maintenance programs or activities should count as failures but you need to be honest about it. Don’t move up a scheduled oil change due to increased vibration readings. With enough equipment history, a good preventive maintenance program should address issues that could causes failures, assuming quality material and craftsmanship are part of the program. Intentionally having unnecessarily frequent preventive maintenance activities can be cheating, too.
Predictive maintenance programs can easily trigger equipment downtimes that avert failures and, in my mind, should be counted as a failure since it a type of damage control. I’d gladly take a 2-day “failure” to correct something then a 2-week failure to replace something.
All these types of programs and measures have very noble intentions. Unfortunately how they are applied and the pressures imparted to get the numbers “right” can distort the entire process and cause a lot of headaches and misinformation. BigInch’s bean counter reference is right…they need to stick with cooking their own books.
Exactly. Right now I have no sense of what JJP means with his 36 hours when trying to relate that statistic to my operations.
MTBF = mean time between failure.
MTBR = mean time between repair
MTTR = mean time to repair
Scheduled out-of-service for routine maintenance, such as a filter replacement or required internal inspection, etc., cannot be counted as repairs, and is never counted as a failure (assuming the air filter did not actually blow apart). Those events would go into establishing the average availability statistic.
The gray areas are what might be defined as significant enough to include or simply be ignored, such as 5 minutes on standby to tighten a bolt would be ignored, whereas 4 hours outage to allow bolts to cool off enough to tighten them might be significant enough to count as a total failure in some cases, even though no replacement was actually made at all.
BigInch-born in the trenches.
MTBF = mean time between failure.
MTBR = mean time between repair
MTTR = mean time to repair
Scheduled out-of-service for routine maintenance, such as a filter replacement or required internal inspection, etc., cannot be counted as repairs, and is never counted as a failure (assuming the air filter did not actually blow apart). Those events would go into establishing the average availability statistic.
The gray areas are what might be defined as significant enough to include or simply be ignored, such as 5 minutes on standby to tighten a bolt would be ignored, whereas 4 hours outage to allow bolts to cool off enough to tighten them might be significant enough to count as a total failure in some cases, even though no replacement was actually made at all.
BigInch
-born in the trenches.For right or wrong, I think a lot of times the MTBF calc is affected more by the reporting capabilites of the Maintenance Management program, and what it can differentiate as to type of work, than what the plant would like to define as a failure.
If craft hours charged is the best flag available, a minimum amount ($ or hours) can sometimes be set to ensure that minor checks and adjustments are not counted.
If craft hours charged is the best flag available, a minimum amount ($ or hours) can sometimes be set to ensure that minor checks and adjustments are not counted.
I hate it when the darn CMMS becomes the tail wagging the dog. It happens too often.
I'm most familiar with SAP (did I spell that right) and in our application we had limited ability to catagorize activities, some but not enough to get by without administrative rules. I've been away from it too long to remember the exact terminology. We tried to push the data entry down to the craftsman level to capture exactly what work was performed and whatever failure analysis opinions or facts were offered. That's a whole different discussion. Anyway, the garbage-in-gargabe-out principal applies in full.
I'm most familiar with SAP (did I spell that right) and in our application we had limited ability to catagorize activities, some but not enough to get by without administrative rules. I've been away from it too long to remember the exact terminology. We tried to push the data entry down to the craftsman level to capture exactly what work was performed and whatever failure analysis opinions or facts were offered. That's a whole different discussion. Anyway, the garbage-in-gargabe-out principal applies in full.
We actually use MTBR rather than MTBF because the word failure is politically incorrect. We do not count any PM work (oil changes, filter changes etc.). As I noted above, it is only if someone believes the equipment is failing (something is wrong) and enters a WO to repair it. Our preventive maintenance tasks come from a different program and are obviously not counted as a failure.
Mechanical seal failures have similar issues. The seal developed a leak. When the pump was on the bench, the mechanic found a problem with the coupling. Someone speculates that the coupling problem caused the seal to fail so they don't count the seal leak as a seal failure. This is too subjective. Instead our criteria for a seal leak are simple: the seal was leaking; the seal was replaced; it was a seal failure.
Any definition has some subjectivity. We have attempted to reduce the subjective nature of the measure and keep it consistent even with changes in personnel, policies or practices. And ultimately, we don't need to compare our measure to any outside standard. We have three refineries that all use the same criteria. We can compare among those plants. And we can compare our performance to our historical performance to see if we are getting better or worse.
By the way, these definitions were not inflicted upon us by the bean-counters. We set the definitions ourselves for the reasons I have mentioned.
Mechanical seal failures have similar issues. The seal developed a leak. When the pump was on the bench, the mechanic found a problem with the coupling. Someone speculates that the coupling problem caused the seal to fail so they don't count the seal leak as a seal failure. This is too subjective. Instead our criteria for a seal leak are simple: the seal was leaking; the seal was replaced; it was a seal failure.
Any definition has some subjectivity. We have attempted to reduce the subjective nature of the measure and keep it consistent even with changes in personnel, policies or practices. And ultimately, we don't need to compare our measure to any outside standard. We have three refineries that all use the same criteria. We can compare among those plants. And we can compare our performance to our historical performance to see if we are getting better or worse.
By the way, these definitions were not inflicted upon us by the bean-counters. We set the definitions ourselves for the reasons I have mentioned.
- Thread starter
- #10
Plenty of food for thought from the replies. We are presently around the 44 month mark. Our problem is that we cannot compare with other refineries in our group because they use different criteria. We count failures as direct interventions caused because the pump fails to function as designed. However seals that "drip" bearings that "rumble" but continue to operate as expected are not counted until the machine is actually taken out of service, ie entering a work order into the system doesn't start the clock running.
- Status
Similar threads
- Question
