Hi Daryl,

quote:

As you say Steve, I don't buy this at all. Reliability is driven by the requirements of the plant, production process, or safety / environmental requirements. Not by the lead time of inventory items.

May I challenge you on this statement which I think is a little incomplete – and as you said, was a quick response…I am sure you know this but readers may miss the subtle points and conditions which are missing from your short response.

The reliability of the plant is driven by its design and the way it is operated….which I think you are getting at. After this, its availability is determined by how you maintain it. How you maintain it is driven by many factors and one of them is the cost and capacity to hold spares vs the cost of inspection / prediction. Other factors include the cost of downtime recovery and the list goes on. This is one factor and I will call this Factor 1.
The other factor is this. To change the spares holding requires analysis, approval and time to restock. In our terminology this would mean a modification. RCM / PMO will find many modifications to put in place and these take time to do and may not get approved. So in this case, the immediate maintenance policy should be to inspect frequently until the spares situation can be addressed. Now you may ask “How often do you find this?” and the answer is rarely, but nevertheless, it illustrates another reason why you should buy my statement.

Back to Factor 1.
If the part that fails is a cheap one and does not deteriorate on the shelf and the cost of doing the inspection is expensive (intrusive even), then it is likely that the maintenance department should stock spares on site. However, if the part is very expensive and the inspection routine is cheap and non intrusive, then the purchase of the part to hold in the store may be rejected on economic grounds. It may be far cheaper to do the frequent inspection and purchase the part when required.

Readers may like to think about how they would organise a spares package for a military exercise or on a merchant ship or at a mine where the barge takes a month to get there. I read Ricky Smith’s comments earlier in this post – he did not say it but I would like to guess that he wanted to know as much about the current state of his equipment as he could because the lead time to spares would not have been a variable in the short term at least. He was prepared to increase the rate of inspection based on many factors including uncertainty of spares supply.
Thanks for raising these points Daryl - they are important
Regards
Steve

This message has been edited. Last edited by: Steve Turner,

quote:

As you say Steve, I don't buy this at all. Reliability is driven by the requirements of the plant, production process, or safety / environmental requirements. Not by the lead time of inventory items.

... One more thing Daryl... the thrust of what you say is certainly right in my opinion... maintenance people need to figure the inherent reliability of the equipment in its given operating conditions and decide what needs to happen from that point. Not start with a given inventory value and try to make that somehow fit the machine and operating context.

Gday Steve,

I think we agrew with the basis of the discussion, but we seem to be having a bit of a smenatical discussion around a few other areas. In the spirit of this I will try to keep on splitting hairs because I think that some of the points are important.

For example:

quote:

The reliability of the plant is driven by its design and the way it is operated….which I think you are getting at. After this, its availability is determined by how you maintain it.

Am not sure I agree wholeheartedly with this one either Steve.

The reliability of the plant is, as you say, a combination of its design and waht we require of it. This in turn will drive the management strategy, which will be a combination of maintenance, operational strategies, and potential modifications.

The reliability isn't only driven by how we operate the plant and how it is contructed and designed, thses are only two factors among several as you point out later in your post.

Reliability and availability are not necesarily linked, high failure rates can occur even with plants that have relatively high availability. Separating these two into the maintenance for availability and design and operations for reliability doesn't do it for me Steve.

quote:

How you maintain it is driven by many factors and one of them is the cost and capacity to hold spares vs the cost of inspection / prediction. Other factors include the cost of downtime recovery and the list goes on. This is one factor and I will call this Factor 1.

I. like yourself, am not too keen to get into all the potential factors driving how we select management policies for pieces of plant. So I will restrict my answer to spares holding.

The cost and capacity to hold spares is not often an issue today. Many companies do still hold spares, this much is true, but many are moving away from this practice. (As much as is possible) vendor held stock, consignment stock, and other methods of shared risk are becoming the norm more than the exception.

Using manufacturers or vendors to bear the brunt of managing lead times, holding costs of spares etcetera. So, in one scenario the cost of holding spares and of meeting lead times is something we may be able to shift onto a suppier for a cost that is very small compared to owned stock management.

However, in the other scenario then yes these are considerations in choosing a maintenance regime. Prior to considering whether a task can be done at the time allowed by our inspection interval we need to make sure that we can have the part available. Holding costs of this item (if they exist) are part of the considerations that we need to run through before deciding that we should or should nor hold the item.

However, in most industries, I have yet to find examples where holding an item is so prohibitive that is makes a company decide not to do a predictive maitnenance task. For example, holding $50,000 tyres within the mining industry is a massive on-cost, but still less than the effects of letting the tyres all go to grief.

Stores management is an art of its own, and getting efficiencies in this area, while still meeting lead time targets, is part of the challenge of this area. I don't actually consider this to be strictly a modification as arising from the RCM / PMO type analyses, rather an area to be analysed on its own with its own parameters. (But with the maitnenance requirements fed in from RCM/PMO style thinking)

quote:

However, if the part is very expensive and the inspection routine is cheap and non intrusive, then the purchase of the part to hold in the store may be rejected on economic grounds. It may be far cheaper to do the frequent inspection and purchase the part when required.

I am not really on the same page here either Steve. rejecting on economic grounds will depend on the impacts of non-availability of the asset, plant or system. Very expensive or not, such as tyres, the obvious costs of downtime in this case makes it the obvious choice. So purchasing on failure, if the lead times are not able to be met, means building an allowed level of downtime into the stores holding strategy. Not bad if it is a small cost of downtime, but pretty serious if it is not.

Cheers,

Hi Daryl..

Topic 1.

quote:

The reliability of the plant is, as you say, a combination of its design and waht we require of it.

Hi Daryl..

First thing Daryl - I said the way we "operate" it not what we "require" of it. These words make a lot of difference.
When you have a piece of machinery and you operate it in a certain way it does not have any idea about what you require of it. If I am a machine I just know that this is the way I am built and this is the way I am being operated and therefore this is how I am going to fail and in what patterns.
You can only change these inherent factors by modification. I, as a machine built a certain way and operated a certain way have a certain inherent reliability – and that is it.
If, however I have smart management, they will take an active interest in my health. They will give me regular checks and on occations remove certain parts because managment knows those parts can only last a certain time etc etc - you know how to derive maintenance programs.
After maintenance, you will not see the inherent reliability - you will see an aparent reliability.
With a certain aparent reliability, availabilty is a factor when you add time to repair. In your discussion, you are talking about apparent reliability after maintenance and I am measuring it as a fundamental of design and operating conditions. Who is right – I don’t know – perhaps someone on this forum can comment.

Regards
Steve

Topic 2

quote:

I have yet to find examples where holding an item is so prohibitive that is makes a company decide not to do a predictive maitnenance task.

Daryl - this is not what I said at all. I said there are times that companies will not carry stock because they have frequent inspections and can get the part in time when required.

Topic 3

quote:

I am not really on the same page here either Steve. rejecting on economic grounds will depend on the impacts of non-availability of the asset, plant or system. Very expensive or not, such as tyres, the obvious costs of downtime in this case makes it the obvious choice. So purchasing on failure, if the lead times are not able to be met, means building an allowed level of downtime into the stores holding strategy. Not bad if it is a small cost of downtime, but pretty serious if it is not.

Cheers,

Economics include the cost of not having the part and the subsequent financial impact on the business. I thought this was a generally accepted terminolgy - my apologies.
Steve

Rui,
Even though you asked Daryl to respond to your post last week, since you talk about the paper I wrote and Daryl is yet to answer, I though I would respond.

quote:

First reason: It doesn’t make sense in my mind to adopt a constant time interval when you are expecting a degradation failure mode to reveal. It should be a progressive decreasing interval instead, as it becomes more and more likely that the failure occurs as time progresses.

I think we are all in agreement that as noticable degradation sets in, the frequency of inspection should increase.

quote:

Second reason: If the consequences are merely operational (safety and environment are no concern), why not to accept a certain level of risk of not detecting in time the failure in progress (some point ahead of point P) and suffer the consequences? If it is only a matter of money, the consequences might not be too expensive as compared to inspections costs over time. It might be worth the risk. Thus, purely economic grounds wise; there may be a calendar of inspections that might return the minimum cost per unit time. Don’t you agree?

Your thinking is right RUI however the mathematics mean that the case you are talking about does not exist according to my maths.
A condition based task is either cost effective or not. First point - so for some tasks there is no economic justification at any inteval whatsoever.
The second point is for tasks that are economic, the mathematical point (interval) of least cost (according to my maths that is) is the PF interval itself. This means for all tasks that are economic, the lease cost interval will be at PF.
Now this assumes that the rectification action is instantaneous. So that is why I suggest that the interval is reduced by a period sufficient to allow rectification.
Now the third assumption that is made in my maths is that the inspection will be 100% reliable in itself, and this is not always true. In cases where inspection reliability is not 100%, it is possible to draw a curve of cost to find the minimum cost based on variable levels of inspection reliability. First of all, I find that if the confidence is high, then there is little impact on the point of minimum cost. Secondly, this equation is dependent on MTBF which is also a variable so one can quickly get caught in mathematics that is compounded by increasing numbers of guesses (such as the cost of one inspection) and purely a waste of time.
The paper I wrote on cost minimisation algorithms highlights some of these assumptions. Cost minimisation algorithms have been popular in the past because, as you have pointed out - it makes a certain amount of sense to us engineers - but in reality some of these algorithms are simply wrong and others make assumptions that are very suspect to say the least.
I would be keen to get your comments - this is my mathematics and I am an engineer not a mathematician.
Steve

Gday Steve,

Rather than go over the same ground again I am going to summarize where I think the conversation has gone to. Otherwise we are going to get into a he said / she said type of discussion and they don't work on posting boards.

I have stated several times that it is the combination of how an asset is designed and what we require of it that determines its reliability. (And availability for that matter)

My point, which I thought I also made previously, is that what we require of it determines both how we operate it and how we maintain it.

This is common thinking surely.


I agree with your comment regarding economic grounds, but your initial comnment implioed something different. (Below) So I answered what I thought you were getting at. Good to see we are in agreement on this one.

quote:

However, if the part is very expensive and the inspection routine is cheap and non intrusive, then the purchase of the part to hold in the store may be rejected on economic grounds.

My original point was, and I am going to expand a little here because I wrote it in a hurry,:

quote:

If you have a P-F Interval of two weeks, and you have taken the decision to do this task based on the effectiveness and applicability criteria within the RCM style thinking, then you need to ensure that you will be able to have the part in time.

Whether this means a vendor held solution, a consignment stock solution, a squirrel store solution, or van stocks, or whatever form you have it available... you need to have it available.

As we said in the beginning the reliability of the plant is not dictated by the lack of immediate availability of the parts.

This is the 21st century and there are a vast number of ways to have the part available to you.

I don't really agree with the diminishing inspection interval thinking either. But lets leave that for another day.

Thanks Daryl,
Discussion on this is over then as far as I can see.
I would like to close off with Rui if he is still interested? I am interested in his point of view since he said it is different to mine.
Regards
Steve

Hi,

I have been out for a few days holiday in the country side (no computer, no mobile... the kind of break every one should have once in a while!). I am sorry for having left a few questions unanswered.

Steve, I am also an engineer but maths (quantitative methods for the decision making) is mandatory in my field of activity as I deal very often with situations where I have to advise managers of the best possible courses of action in environments of uncertainty.

My viewpoint on the two (interrelated) issues which are being discussed in this thread (1. calendar of inspections to prevent evident failures and 2. keep or not spares in stock) is that they are clearly objective and quantifiable on economic grounds. Any particular case can be easily equated and an acceptable solution can then be found and implemented.

These two subjects (along with others in the field of maintainability) have been my concern for some time a few years ago. I did some extensive research at the time (that is when I found Steve’s article on the P-F interval in the WWW) and I came up with algorithms of my own which I proposed in my doctoral thesis and presented in seminars. Later on, non-commercial software was developed and the algorithms were implemented in a few big industrial companies and returned satisfactory results so far.

I chose to attach herewith copies of two original documents of mine treating these subjects. Despite the documents are written in Portuguese, I think they are self-explanatory in what the mathematics is concerned. It would be a tremendous effort to put all the texts in English.

Allow me to describe in a few words the two cases treated in the attached papers.

PF interval:

Consider a mechanical device pertaining to a piece of equipment.

1. Predominant failure mode described by a Weibull distribution with shape parameter of 2, a scale parameter of 8,000 hours and a location parameter of 0 (wear out typically);
2. Accumulated life time: 3,000 hours;
3. Last inspection: 2,600 hours ago and presented no sign of the onset of a failure;
4. Time remaining before the next overhaul: 9,000 hours;
5. PF interval: 500 hours (it could be considered to follow some probability distribution);
6. MF interval: 50 hours (the minimum time span you have to react and do something to prevent the functional failure – point F – meaning, therefore, that you have a time span of 500 – 50 = 450 hours to detect the failure in progress, if you are willing to prevent it from happen);
7. Cost of an inspection: 4,000 €;
8. Cost of the consequences of a failure: 100,000 €;
9. Labour and materials for the repair: 3,000 €;
10. Cost of capital: 25% per year;
11. Working regimen: 8,760 hours/year.

As can be seen in the article, I used the concept of “safe” and “unsafe” time windows and conditional probabilities of failure given that the onset of a failure was not detected at each inspection. I used both an analytical method and Monte Carlo simulation and concluded that the optimum reliability between inspections is 0.91. The expected number of inspections to be accomplished is 22 over a period of 9,000 hours, unless a failure occurs in the mean time.

Note that, in the case of R = 0.90, intervals between inspections 7, 8, 9 up to 20 are shorter than the PF interval. Therefore, I considered that the onset of a failure would be detected for sure if it would happen. On the contrary, prior to the 7th inspection, the onset of a failure presents a certain chance of not being detected in time to be prevented.

This approach is valid to start with an inspection program but it should be adjusted every time an inspection takes place and nothing wrong is noticed (from the Bayes theorem). The article treats this issue too.

The software accounts for a few more variables.

Spare parts in stock

Consider a part that is being questioned whether it should be bought and kept in stock or ordered only when it becomes actually needed and the following data:

Remaining life of the equipment: 3 years;
Working regimen: 1,850 hours/year;
Mean rate of failure: 0.00025 failures/hour;
Unit cost of the part: 7,000 €;
Opportunity cost (cost of lost production): 3,000 €;
Capital cost: 15% per year;
Warehouse keeping cost: 10% per year

As can be seen in the attached article, the expected cost of the lost production is 2,678 € while the expected holding cost of the part is higher and equal to 3,390 €. Therefore, the purchase of the part is an alternative to be rejected.

Once more, please apologise for the explanations in English to be so scarce. My intention is just to illustrate quantitative methods that can (and should) be used by maintenance pratictioners more often.

Regards,

Rui

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

Inspections_and_Spare_1.zip (251 Kb, 24 downloads)

Thanks Rui,
This will test my maths.
Consequently I will need a little time to meet the challenge.
Good to see you had a break. I need one too.
Regards
Steve

Hi Steve,

I am sorry but I ziped a different version from the one I intended to of the document entitled "Capítulo II". Please download this new version "Inspections and Spare_1" (I replaced it in my previous post too).

Thanks.

Rui

Inspections_and_Spare_1.zip (251 Kb, 25 downloads)

Rui / Steve,

The first series of questions in the applicability criteria of the SAE 1012 diagram 17 asks the following questions: (I am sure there is a reference to it in the standard also, but don't have a copy to hand)

1. Is there a P-F Interval?
2. Is it consistent?

I would suggest that if we are using an RCM style approach then diminishing frequencies due to diminishing P-F intervals is probably not the way to do things.

Cheers,

Hi Daryl,

It is true that SAE 1012 says (page 30, point 13.1.c) “The task interval shall be less than the shortest likely PF interval.” The decision diagram (page 49, figure 17) also refers the PF interval the way you stated in your post.

But my view point is that SAE 1012 is just a practice recommendation – not something to be done compulsory. I don’t agree with its terms with regard to PF interval. I gave the subject some extensive thought a few years ago and developed the approach described in my last post. Perhaps some peers don’t agree with it. In this case, I am minded open and would like to hear their reasons explained in solid mathematical grounds. There would be no progress if one keeps aligned with rules. Their validity has to be questioned from time to time in light of new contexts or knowledge advances. Don’t you agree?

Rui

Rui tudo bem!

I couldn't agree more with both points.

1. SAE 1012 is the guide to the standard, a good practice guide only. Is there specific guidance within the standard itself as to P-F interval setting? (I don't have my copy with me)

2. Yes, things definitely need to be challenged.

Having said that I don't agree with the diminishing P-F strategy you set out. Happy to comment but it won't be today.. sorry mate!!

Hi all -
My statement about decreasing intervals of PF had the condition - after noticeable deterioration had set in. Now for condition monitoring, using one example - if we find a crack in a structure, then the rate of crack growth will increase depending on the crack length itself. This means the PF is reducing with time hence, there are two options; increase the rate of inspection or repair the crack before it goes "bang". Now all aeroplanes have cracks Daryl - not all of them have been repaired... You know this so I suspect I what I have written has not been written clearly enough.
Regarding constand PF's as per your Point 2 "is it (the PF Interval) consistent.
Leaks (for example) are generally variable in their rate of degradation... but almost every maintenance analysis I have done has had leak checks. To do otherwise would mean we would not be asked to do any further analyses for the clients and we would be out of business.
I have not had a chance to review Rui's formulae - suffice to say that if they are saying that as equipment gets older the inspection rates should be reduced over time due to equipment age, then I would like to know the rationale. Most people who have read Nowlan and Heap take the approach that age and deterioration rates from initial initiation are not closely linked in most cases... or at least that the PF may reduce for some failure modes if the equipment is older, but this would not take the deterioration rate out of the same order of magnitude (weeks to less than a day for example).

Steve,

I haven't read what you wrote previously. I was just speaking to the general point. Apologies for any confusion.

Am trying to get back to this one in the very near future.

To avoid any misunderstandings, I would like to make it clear that, in the model I created, the decreasing intervals of PF are valid until the onset of a failure is noticed, even if the time intervals are shorter than the PM (a fraction of the PF interval which is the minimum time still left to react and prevent a functional failure - the so called safe time windows). When, as the result of an inspection, a sign of a failure in progress is noticed, the equipment is scheduled to be stopped asap (within an estimated period before the point F is reached) and the deficient part replaced or repaired.

Regards,

Rui

Dear Rui, I'm trying to understand your point.

Do you mean we should decrease the time between inspections before or after the onset of potential failure (point P)?

Hi Josh,

As you know, the probability of a degradation failure mode (erosion for instance) to occur increases with age, therefore it is logic to consider that the frequency of inspections must be increased or, in other words, the time between every two inspections must decrease progressively. When the onset of a failure is noticed, then, it is up to you whether to stop immediately or proceed based on the (eventual) knowledge of the PM interval (rather the PF interval). Within this period, you still can perform a few more inspections, just to be sure of the rate at which the failure progresses, but I didn´t consider these eventual extra inspections in my model.

If the failure that you are expecting is a random failure mode, then, the time interval between inspections must be constant. But if it is an early failure mode, the time interval between inspections must increase progressively.

Regards,
Rui