quote:

Originally posted by Josh:
Rui, in a plant environment, instruments are normally not repairable so replacement is a common policy eg transmitters. Hoever, control valve trims can be replaced.

Yes Josh, I quite agree. I correct myself: Even if that kind of devices are not repaired and are simply scrapped and replaced by new ones, it still takes time to do it. Therefore a TTR (Time To Replace) always applies after a failure, meaning that indicators MTTF, MTTR and MTBF are still possible to control as time goes by.

I think the term "repairable" is not the most appropriate and should be abandoned as in the end, every system can be "repaired" or "recovered". I would suggest "recoverable" instead. The only distinction, in my view, is to be able to do it while in service or not as I exampled in my last post.

Regards,

Rui

"Recoverable" is more general in scope and would embrace both the terms "repairable" and "replaceable". Misunderstandings would be almost impossible.

What do you think?

The IMechE document is The reliability of mechanical systems by J Davidson under IMechE Guides for the Process Industries ISBN 0 85298 675 0.

Rui, Rolly,

Rui, your statement

quote:

MTBF means Mean Time Between Failures and is the same as "the average time between two successive failures", therefore comprising the time to repair every failure, that is MTBF = MTTF + MTTR (Mean Time To Recover, Replace or Repair).
and
...the missile and the satellite are "repairable" while they stay at the stores but they are not as soon as they are "in service". The batteries are not repairable in service, that is, while the maintenance person in charge is away,

is not in line with the authorities quoted below
1. The IMechE document "The reliability of mechanical systems by J Davidson under IMechE Guides for the Process Industries ISBN 0 85298 888 8 (ISBN differs from the one quoted by Josh), on page 34, says, Quote
Similarly there is a difference between the MTTF of a non-repairableitem and the MTTF of a repairableitem (emphasis mine).
The term repairable has a technical meaning and is not the same as in common English; items such as ball bearings and light bulbs or microchips are termed non-repairable. Also all items subject to hidden failures. Your examples of the battery and the missile do not meet the technical definition of 'repairable'.
2. The book "Reliability Maintenance and Logistic Support by Kumar,Crocker, Knezevic and Al-Haram, Kluwer, ISBN 0 412 84240 8" says on page 76, quote
MTTF is.....a measure of reliability of non-repairableitems... (emphasis by authors. On page 85, they say, quote
MTBF is.... a reliability measure for repairablesystems (emphasis mine).
On page 87, they say, quote
The value of MTBF is equal to MTTF if after each repairthe system is as good as new.

Two points;
a) MTTR has nothing whatsoever to do with reliability; it only affects availability. So it does not affect MTBF, see item (b0 below.
b) MTBF is not MTTF+ MTTR as incorrectly stated in your post.

In computing MTBF, we do only take the time available in service, excluding any downtime for repair.
In computing MTTF, replacement time is usually negligible, so calendar time is considered as operating time. This is arguable, but is acceptable in practical terms.

Rolly, you said
MTBF and MTTF are not the same,
I agree entirely with you; however their numerical valuesare computed using the same formula.

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

Thanks Josh for the information. I looked through the IMechE site and found:

Davidson J and Hunsley C (Editors), The Reliability of Mechanical Systems, 2nd Ed, J Wiley Publishing, IMechE, London 1994, ISBN 0 85298 881 8.

This is a guide for practicing engineers. It provides an introduction to the application of the methods and techniques for assessing the reliability of mechanical components, equipment, plant systems and subsystems.

The following subjects are discussed:

 The philosophy and concepts of reliability engineering
 Analysis of in service reliability experience
 A basic approach to reliability assessment for mechanical process systems
 Techniques for process plant reliability assessment
 Collection and processing of reliability data
 Case studies.

The subjects sound quite appealing to me. Is this the book that you meant Josh?

The title of one of the subjects called particularly my attention "Analysis of in service reliability experience". Davidson also makes use of the term "in service" as I do.

Despite I don't have the book, I have got many others on Reliability and Maintainability and they all refer to "repairable" and "non repairable" systems the usual way that causes confusion. That is why I proposed the term "recoverable" instead. Anyway, here is an example which illustrates what I mean.

Suppose a system composed by two elements (1) and (2). The duty element is at the present moment number (1) and the stand-by element the number (2). It could be a duty pump and a stand-by pump. Suppose that they are alike in what refers reliability and that any one of them fails randomly every 200 hours in average and still that the mean time to repair is 10 hours. Now suppose two different situations:

1. The system can be serviced at any time, meaning that as soon the duty element (1) fails, the stand-by one (2) takes over and the former is immediately taken to the repair shop to be repaired. As soon as number (1) is fully recovered it is put back into place and becomes the stand-by element in turn (a 50-50% arrangement). The system would only fail if the element (2) failed while element (1) was being serviced. This is what is usually known by a repairable system. I would rather call it "repairable (while) in service";
2. The system cannot be serviced at any time, meaning that as soon the duty element (1) fails, the stand-by one (2) takes over and the former stays indefinitely in a failed state. As soon as number (2) also fails, the system fails and only then both elements will receive service (the mean repair time will be then 2 x 10 = hours). This is what is usually known by a non-repairable system. I would rather call it "not repairable (while) in service";

To assess the performance of the two situations, you create a model of the system and use Monte-Carlo simulation to see what will happen if the model is run for a long time (long enough to accommodate a few hundreds of failures).

I run the model for a while and found the following results:

Repairable system (while in service):

Reliability for a mission of 100 hours: 0,9598;
MTTF = 4,610 hours;
Availability = 99.88% (which coincides with the theoretical value).

Non-repairable system (while in service):

Reliability for a mission of 100 hours: 0,9081 (which coincides with the theoretical value);
MTTF = 400 hours (which coincides with the theoretical value);
Availability = 95.21%.

Please note the difference in what MTTF (huge) and availability are concerned.

In conclusion, the two situations are far different.

Regards,

Rui

Vee,

Thanks for your observations.

The difference between MTTF and MTBF seems more like a semantic problem to me. I am aware that the way I interpret those two terms is not in line with the authorities referred by you. I know of a lot more authorities that adopt the very same definition but I also know others who adopt the same definition as mine. I can't see any reasonable reason to maintain such a cumbersome definition: meaning the same but applying to different systems in what maintainability is concerned. It gives rise to confusions and I face them frequently coming from practitioners. I feel quite comfortable since long with my definition (the same as many others) and I have already explained how I easily distinguish between a so-called repairable system (repairable in service) and a non-repairable system (not repairable in service). This is the way I teach my students, write in my books and presentations and advise practitioners – no contests so far! . Everything is clear and there is no room for confusions.

Regards,

Rui

Rui,

quote:

I also know others who adopt the same definition as mine. I can't see any reasonable reason to maintain such a cumbersome definition:

Cumbersome or otherwise, the equation MTBF = MTTF + MTTR is not correct, as far as I know. MTTR has absolutely nothing to do with reliability, it is a completely indepemndant parameter. I hope we can agree on this at least.
Perhaps you can quote the authorities who use your definition, for my reference.

Vee, do you consider an industrial air compressor as a repairable or non-repairable system?

Rui

Rui,
I am answering using the technical definition of 'repairable' not the common English meaning.
Most of the compressor is repairable, but not, e.g., its safety relief valve, because that has a hidden failure mode (fail to lift on demand)
Now you have to tell me why you ask this question without addressing my earlier question.

Vee,

I briefly reviewed some of the English literature on Reliability that I have and found:

Time To Failure (TTF), means "time interval between two successive failures" and is an acronym which is historically applied to components which are not repairable, such as lamps and bearings. Time Between Failures (TBF), on the other hand, means exactly the same but is only applied to components which are repairable.

A side note: What to call a fun blade which is considered non-repairable if I decide, for some reason, to restore it by adding some eroded material by using a suitable and available technology? I remember once in a merchant ship a few years ago during an emergency, having had to restore the outer ring of a ball bearing mounted on a gear box as we hadn't any spares available aboard. Should I have called the bearing a repairable or a non-repairable item at the time? That depends on the circumstance, doesn't it? So why to keep this cumbersome definition which is the cause of so much confusion?

See on this purpose the image attached of an Excel application authored by Paul Barringer where he, himself, uses MTTF for the components and also for the compressor. Shouldn't he had used the term MTBF instead? As you said, the components (with a few exceptions) and the compressor are repairable.

TTF is translated into Portuguese as "tempo de bom funcionamento" or "operating useful time" and TBF as "tempo entre duas falhas consecutivas" or "time between two successive failures". TBF, the way it is translated, comprehends necessarily the time to repair (recover or replace) TTR. For me, the (historic) convention doesn't hold but rather what the words mean. In plain Portuguese TBF = TTF + TTR.

In result of our conversation, I am thinking seriously of replacing the two English acronyms TTF and TBF (the ones that have different meanings depending on whether you are speaking "technical" or plain English) in all I write from now on, by two others with real meaning in the Portuguese idiom. Thanks for the stimuli.

Regards,

Rui

This message has been edited. Last edited by: <Rui Assis>,

Air_compressor.JPG (169 Kb, 12 downloads)

Dear RUI, VEE and all,

The discussion on this thread is getting deeper and deeper. First I would like to clarify my point on MTBF and MTTF.


In reality, machines do not fail as most of us usually try to say. Hey guys, this machine is failed go get me a mechanic. What is right is that a component, part or spare on a machine had failed. And so this is where I believe that repairable (MTBF) and non-repairable (MTTF) had been derived to distinguish one indicator from another. Well of course Im disregarding in this thread decommisioning the equipment !!!

It all started out in one of the company's I am teaching since, I am accustomed in using MTBF all the time, and another consultant who also teaching in that same industry was using MTTF all the time and people got confused if they are the same or different and so I need to make that distinction or otherwise one of us will loose our job of teaching.

If we speak about the machine as a whole then the correct term to use will be MTBF since machines can be repaired, right ? Now if a bulb, bearing, seal, capacitor, had failed and it cannot be repaired but replace then we speak about MTTF. But most of us still are accustomed in using MTBF of the bearing, MTBF of the light bulb and so on.

Vee, I have to go with RUI that when you consider the MTBF of the component, this is the mean time between failure, and we need to consider the repair time that took place to be able to get the machine running once again. While when we speak about MTTF, which is the mean time to fail then we speak about the actual time when the part had reached its last second to operate. Hence, I will go for the conclusion that MTBF = MTTF + MTTR.

I'll speak about availablity & utilization later on and problem with industries that there is no standard formula on this most specially when your equipment had not run or is not loaded.


Cheers,

Rolly Angeles

MTBF-MTTF.ppt (30 Kb, 29 downloads)

Rolly and others,
Let us start at the beginning. When we repair any component, we can bring it to a condition where its reliability at that point is 100%. We call this condition As Good As New or AGAN. As you know, with use, the component reliability falls, untill it fails. Now, repairs done in the field are invariably to a lower standard, partly because of access, lack of skills, special equipment or tools, lighting, dust etc. In such cases the component starts its new lease of life, not at 100%, only but some value lower than this, as if it had already been in service for some time. In extreme cases, the spare component is not available, but to keep the machine running, we may clean and dress up the old component, and bring it to a state it was in just before its failure. In this case, we call the rapair type As Bad As Old or ABAO. You do not expect the component to last more than a few days or weeks, but it gives you breathing time to get a new spare.

For all items that start each new run in AGAN condition, we compute MTTF.
For all items that start some of the runs in non-AGAN condition, the measure is MTBF.

In both cases, the formula we use is exactly the same, the difference is ONLY the fact that in the second case all the runs do not start with AGAN parts, i.e they did not have a reliability of 100% at the start of each run. THAT is the logic to use.

Coming to your complex machine, you are absolutely right. When a machine fails, in fact one of its components has failed. If we replace it with new ones each time, we can measure the component's MTTF. But the machine has, at that stage, many other worn components, whose reliability is less than 100%. So when we replace one or two components, the machine as a whole is still far from AGAN
when we restart it. So the applicable measure for the machine itself is MTBF.

For both MTTF and MTBF, we use the formula:
(Cumulative time available to operate)/(Cumulative number of failures), where
Time available to operate= Operating time+standby time
thus time to repair is ALWAYS excluded from this formula.

Time to repair is an independant variable, that has NO effect on reliability. Since MTBF and MTTF are strictly reliability measures, time to repair is not an issue when we compute them. It matters only when we compute availability.

Rui, as a point of clarification, we use the term 'repairable' to denote any component that does not come back to AGAN condition at the start of each run. Items that have to be replaced with new or AGAN components are termed 'non-repairable'. It does not mean they cannot be repaired, only that they are always repaired to AGAN condition. As one well versed in Reliability Engineering, this definition may appeal to you, as it means the same in English, Portugese or other languages.

For most of us practitioners, this distinction between MTTF and MTBF is not very important. For all practical purposes there is no difference, and not worth losing sleep on.

Rui, that's the book albeit ifenret ISBN no.

Just curious, have you done reliability modelling and simulation for industrial plants before?

Yes Josh, I use simulation often to model and anticipate the behaviour of industrial systems.

Regards,

Rui

Vee,

The process that you refer is still more complicated and known by "minimal repair process" which, as you know, deals with the aging of a system which is maintained by replacing or restoring only a small percentage of its components. This will leave the system in approximately the same state (age) it was in just prior to each failure. To model this point process, we define an intensity function as the rate of change of the expected number of failures with respect to time which is also referred to as the renewal rate, failure intensity, peril rate, or the Rate Of oCurrece of Failures (ROCOF). This intensity function, which is an unconditional probability of failure within a certain interval, is often (again) confused with the hazard rate function, which is a conditional probability of failure within a certain interval given that the unit has survived to the beginning of the interval. The hazard rate function is a relative rate pertaining only to the first failure, whereas the intensity function is an absolute rate of failure for repairable systems.

One more cause of confusion...

In this case you case, you don't have a single MTBF but rather an interval MTBF which diminishes as the equipment ages.

Another cause of confusion...

Of course that MTBF has nothing to do with reliability but rather to maintainability. MTBF only influences availability. I agree.

This activity of ours is definitively a complicated one...

Regards,

Rui

Thought I'd share a few other points of view on MTBF

Attached is an article from RELIABILITY®Magazine on Equations of RCM that contains a definition of MTBF

Also a link to wikipedia and what they say along with a link to an article cited in their backup information
http://en.wikipedia.org/wiki/MTTF

http://www.vicr.com/documents/quality/Rel_MTBF.pdf

I think this points to one of the main problems we have in our profession, too many definitions for the same terms. How can we all talk about something when we always have to define the term because we've modified it to fit our particular company's environment, usually to make someone look better than they really are.

I think that OEE and availability should only be measured over a one year period of time (except for looking for improvements and trends) using 24 hours 7 days a week because that's the way we run our businesses.

Just becausse you're not running 24/7 now for whatever reasons doesn't matter. OEE and availability should provide info on how much capacity you have and how much it could be improved or utilized if changes were made.

RCM-Equations2.pdf (1,945 Kb, 25 downloads) RCM Equations

Joe,
Thanks for the two references.
The wikipedia one says, and I quote
Reliability is quantified as MTBF (Mean Time Between Failures) for repairable product
and MTTF (Mean Time To Failure) for non-repairable product

Rui,

quote:

The process that you refer is still more complicated

I regret I have to disagree with you. One can make anything as complicated as one wants, but I have outlined some basic principles, namely,
If an item starts every run in as good as new or AGAN condition, we compute MTTF.
If an item starts some runs with less than aGAN, we compute MTBF.
For both we use the same formula, i.e. time available to operate/number of failure events.
MTTR has absolutely nothing to do with reliability and hence MTBF.
Items that start each run with AGAN condition are termed non-repairable (even if they can in practice be repaired)
Items that start some runs with less than AGAN are called repairable.

I am not sure what is complicated in these statements. We dont need advanced maths to explain any of these to people.

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

Rui,
We should normally use the terms Failure Rate and MTBF in the constant hazard rate context(see Davidson, The Reliability of Mechanical Systems, ISBN 08529 8818). In all other cases Rate of Occurance of Failure is preferable (see Hoyland A, M. Rausand, System Reliability Theory, Wiley, ISBN 0471 593974.
Following this practive. MTBF does not vary with age.

Terry:

Wow! Lots of messages back and forth on simple subjects! Thanks for keeping me posted.

For the differences between MTBF & MTTF see: http://www.barringer1.com/TWMS.htm#MTBF_MTTF.

You can find tons of definitions and examples in MIL-HDBK-338 at http://www.barringer1.com/mil.htm

or specifically at http://www.barringer1.com/mil_files/MIL-HDBK-338.pdf

Most folks GET IT concerning availability. Unfortunately they quickly start to believe availability is THE answer to everything. Fewer folks GET IT concerning reliability.

To help people understand UNreliability (and it's complement, reliability) I'm thinking about offering a no-charge device that will Velcro over an automobile's master computer (on cars with a left side driver it's on the passenger side inside the cabin and near your ankle).

The GET IT device has a random clock. Once a day the clock will disable the computer for up to a 15 second interval. The "repair" can be accomplished by simply opening the hood and slamming it shut hard enough to "fix" the automobile and reset the clock for the next day. The short repair time will not significantly detract from availability of the automobile. But very quickly the driver will love to hate the repairs!

Perhaps the random failure of the automobile's computer will give the driver an appreciation of UNreliability and it's complement reliability. The key to reliability is to prevent failures. You cannot repair yourself to happiness.

I've published a couple of articles to my website that you may find of interest. See posted papers 35 & 36 at http://www.barringer1.com/Papers.htm . You'll like the SAP data in paper 35 and the corrosion data in paper 36.

----

Paul Barringer
Barringer & Associates, Inc.
Phone: 281-852-6810
FAX: 281-852-3749
Email: hpaul@barringer1.com
Web: http://www.barringer1.com