I frequently find companies maintenance departments that do PM and yet manage spare parts without distinguishing demand due to wear out failures from demand due to random failures – the only fraction of demand that justifies holding permanently some parts in stock. Demand is mixed all together and the average stock and the holding costs raise consequently to levels higher than they should be.

I would like to know your experience on this issue and to hear your comments.

Rui

How to distinguish demand for spare parts due to wear out failures from random failures? Any criteria or guidelines?

Are you saying the only fraction of demand that justifies stock is demand due to random failures?

Hi Josh,

Yes, definitively. Parts must be ordered just-in-time for PM actions and just a small number must be kept in inventory just-in-case a random failure occurs. Don´t you agree?

Rui

AMAP (As much as possible), we like JIT but I think this idea of getting spare parts when needed is not practical for small qunatities of spare parts. We have to stock some materials especially if the location is far from suppliers. Probably the most we can do is to have a price agreement with the vendor which can stock the parts if he has several customers for fast delivery and to avoid multiple quotations.

For just in case (JIC) for random failures, is it required to stock? Isn't the idea of price agreement applicable here?

Wear out spare parts may be easily identifiable eg gaskets, impeller, wear ring, bearing, lube oils etc.

But how to easily identify which components undergo random failures? What I understand complex equipment or system or assembly undergo random failures due to various failure modes. So which parts to stock? Another problem is I don't think many have plotted the RCM curves to know which equipment fail randomly in terms of timing.

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

I caution that I am not an expert with spare parts or with RCM. But I do like to ask a lot of questions.

Do you stock your complete industrial plant or do you stock a portion of your industrial plant? Do you stock assemblies or do you stock individual parts?

I do not have the answer for all these questions, but I think using an RCM type analysis may provide a starting point. Is RCM a way to understand the consequence of a "*failure" at an industrial site? Is it a way to set site equipment maintenance strategies? If equipment is not on a maintenance strategy it could be very difficult to discern what needs to be stocked as spares.

*Failure seems to be such a cloudy term – I no longer see it as black and white, but grey. Failure could mean a specific part, assembly, equipment, or system failing (it almost makes me thinks of cells, tissues, organs, system, etc.)

Anyhow back to the discussion:

For example:
In the RCM analysis there is some understanding of the consequence if a component fails. Is there a health, safety, environmental, company image, "work place" cultural, production, or other type of impact? For online spared equipment (Main/Standby, 90:10, 50:50) impacts may point to increased risk of losing system availability, but little physical consequence.


Once the consequence of failure is understood then it is necessary to understand how failure could happen and how often it could happen. What are the dominant failure modes to the component failing? What are the contributing factors to these dominant failure modes? You could pull out your crystal ball, but I find trades-people and operators are just as good if not better at helping create some sort of Fault Tree Analysis. With the fault tree there should be some idea of probability of a particular failure path occurring.

So now we know:
· What the consequence of failure is
· How it happens
· How often it happens

If industry benchmarking is correct we know very few failures are time based. Most failures are not time dependent, leaving us with condition based monitoring or run to failure strategies. Again individual sites will make these individual choices.

If most failures are not time based then the timeframe of when most parts are replaced is not known. We must try to determine a condition based task that will allow us to know when a failure will occur. With condition based monitoring forecasting may not provide enough lead time to order the part. There are a number of factors to consider with forecasting – how well is the monitoring performed? How clear are the criteria? Are the Criteria correct? Confidence in the workforce acting upon a found condition? If the condition is missed how bad does the failure become? What other parts will be required to fix this more "significant" failure.

Note: I could go on with more questions but I prefer to summarize at this point.

In summary, without some idea of what can fail, how it fails, how often it could fail, and what the consequences are, it is very difficult to state what parts are necessary to stock. Therefore creating a maintenance strategy based on possible failure routes, probabilities of occurrence, and monitoring these routes is crucial to identifying parts for stock and parts for order on request.

Such nebulous discussion is difficult to describe in text.


I will try and add an example in my next post.

Example:

A reciprocating gas compressor

Originally a compressor was installed and operated for 2 years with minor maintenance activities occurring (oil, lube, filters).

A realization was made that the valves for the compressor could wear out – so they were replaced on a time basis. And every year and a half later the valves were replaced.

The compressor operated fault free for over five years. 2,000 hours after the last valve replacement a catastrophic compressor failure occurred. It was determined that a 1st stage discharge valve had fouled. The valve did not fully close and this allowed for heat build up in one of the chambers. The piston in this chamber expanded and contacted the bore. The piston and bore were scoured from the contact and the increased friction of the contact created high stress in the crankshaft of the compressor. Upon noticing the sound and vibration from the compressor a millwright had it shut down.

The consequence of the compressor shutdown resulted in inefficient use of raw materials. The fault tree of the compressor failure indicated that the valves failures had a mediocre time related failure characteristic as many contributing factors played a role (Rebuild, install, and process factors affected the probability of failure). Therefore condition based monitoring and age replacements were both implemented.

*As a side note replacing the valves on condition is the least maintenance cost strategy, but it is known the risk of failure does slightly increase with time. There are other components within this system that are time-based failures. The optimum least cost business strategy is to replace these time-based parts and valves at the 1 and ½ year mark. The expectation of changing the valves on condition is once every 5 years.

We now used our best information at our disposal and estimated that we would have 5 to 40 days notice of failure (assume 99% of the time). The lead-time on the parts are 14 to 21 days and may not include the time to quote, plus the manufacturer shuts down for two weeks in the summer. We could shut the equipment down without allowing it to go catastrophic, but we would lose 20,000/day. The parts cost 15,000. So calculated our risk of not having the parts in time and the administration of chasing this issue as an emergency and decided to stock the parts.

We felt there is a small chance we may miss the condition or not act in time (assume 1% of the time) and the compressor would go catastrophic again, but we see this as a once every 5 years multiplied by 1% = 1 in 500 year occurrence and if we held a crankshaft, pistons, etc. for a 1 in 500 year event we were not optimizing the use of our money.

I would not suggest awaiting for a failure to occur prior to assessing equipment for maintenance strategies, spare part strategies, and online spares operational strategies. And all of this starts with understanding "what is the consequence of failure?".

Rui,

I think that your original question related to an increase in inventory holdings that results by not recognising the difference between parts used for random failure and those used in a timebased manner. Is that correct?

If so, then in my view the answer lies in how much extra is invested and so is it worth the time and effort to try and solve the problem? (Just to be clear I am not talking about deciding between planned or run to failure strategies but separating the spare parts usage.)

If you are talking about items that don't cost much then it may not be worth the bother.

The solution lies in a case by case review of the extra investment and the effort required to reduce that investment.

I should also mention that I am probably biased as I developed a process for reveiwing MRO inventory investment and reducing that investment without impacting operational capability. You can learn more about this by visiting www.InitiateAction.com/inventory_management.htm

Let me know if I got your question right.

This is the way I deal with this issue and I feel quite comfortable with it so far.

Josh an Jaz,

As you know, working orders from a CMMS are normally registered as having been originated in the production department (asking for a CM intervention) or in the maintenance department (PM based on running time, calendar time or results of inspections PdM). Parts that have been replaced as a consequence of any of these actions are entered in the DB and stats can be built up at any time in order to distinguish the demand from CM or PM. Nevertheless, parts that have been found deficient and replaced after an inspection must be obviously included in the CM group. The discharge valves from the reciprocating gas compressor that use to fail once every 5 years, as per the case Jaz described, should undoubtedly enter this group.

So, for any one single item an average of X units/day originated in CM + PdM and updated on a timely basis, will always be available, allowing you to determine the ordering point at any time, based on that level of demand, lead-time (time span between the moment you notice the need of a part and the moment you get it in your hands free from any administrative constraints what so ever) and level of service. This fraction of demand is typically represented by a Poisson probability distribution, which makes the calculation of the reorder point an easy task. Of course, there are circumstances when the unit price is too high and you have to trade off holding costs against opportunity costs.

The order quantity can be any number that you like, feel more comfortable with, a minimum imposed by the vendor or, still, an economic quantity given by the well known Wilson expression, rounded to the nearest multiple of a standard box content.

This ends the just-in-case picture according to my view point.

The fraction needed for PM actions, based on accumulated life time or on elapsed (calendar) time, will be ordered whenever needed, by activating the MRP (Materials Requirements Planning) algorithm in the CMMS.

This ends the just-in-time picture according to my view point.

I understand that some times it is difficult to say definitively how a certain use must be classified but it doesn't void the merit of using a scientific method instead of rules of thumb, don't you agree?

Phill,

I think that I have answered your question. Thanks for your view point on economics and the link, which I will visit with curiousity and interest. Would you like to comment the few lines that I dropped here?

Regards,

Rui

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

The case of the discharge valves from the reciprocating gas compressor described by Jaz raises the issue of being or not worth while keeping expensive items in stock. The crankshaft for instance, would it be less expensive, on an annual basis, to keep one unit in stock or none and run the risk of the one in service to be badly damaged in result of a failure (inherent or caused by other parts) some time in the future?

If Jaz doesn't mind, I would very much appreciate to work with real numbers. All the information I need is the following:

Unit purchase cost in $/unit;
Lost production cost (opportunity cost) in $/hour;
Corporate cost of capital in % ($ per each 100$ immobilized during one year);
Warehouse cost in % ($ necessary to keep 100$ of material in stock during one year);
Failure rate (due to inherent failures plus due to failures caused by other parts which the crankshaft interacts with);
Mean working regimen in running hours per year;
Useful life of the crankshaft (which is most likely the same as the air compressor itself)

As a side note, it can be the crankshaft or any other part you wish.

Rui

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

It really come down to what is the cost and effect of the failure to your processes.
If you can afford to have the euqipment down for a while and have a strong planning and scheduling program, you should then concentrate your effort in preventing or predicting the failure rather than planning the just in case scenario....

Rui Assis

This is my attempt to provide as much information as possible. The following numbers are not completely accurate, but I believe they are reasonable.

I will start by describing more of the production impacts of stopping the compressor.
When the compressor stops there is limited lost production perhaps 0.5 to 1% of 850 tons for the duration of the outage and for 7 days after the compressor is returned to service. A ton sells for $1100, and currently costs $700 to produce. Most of the costs are variable costs based on raw material costs. So if we take a partial hit there are very few fixed costs being paid so the upper tons still have a high production cost of $600/ton.

Unit Purchase Cost – I am unsure if this is the cost of purchasing a production ton or the cost of the part(s).
Lost production cost – I am unsure if this is the cost of missing the large margin tons at the top after the fixed costs have been covered
Corporate cost of capital, assume 12%
Warehouse cost, assume 25% (I have heard variations, but again this will be a mixed bag of fixed costs and variable costs averaged across all spare parts)
Failure rate of the valves, assume failure pattern of a shallow bathtub with an Eta of approx. 2 years.
Failure rate of the crankshaft, assume a failure pattern of constant to slow ageing with an Eta of approx. 20 to 25 years.
The compressor runs fulltime 24hours/7days a week.
Assume the expected life of the compressor is 40 to 50 years with obsolescence of the compressor or business as the main drivers for stoppage.
Lead times on compressor parts is expressed in the original example – 14 to 21 days with a possible 2 week summer shutdown of the supplier.

The part costs for a crankshaft failure: 60,000 in parts (Crankshaft and ancillary parts).
The valves are an additional 15,000 (and would probably be replaced at the same time).

The biggest contributing factors to the crankshaft failures would be: valve failure (highly likely if not monitored 1/5 yrs.), oil condition (highly likely if oil, filters, and seals not maintained), bearing condition for the crankshaft (likely if bearing condition not monitored – instrumented vibration and physical measurements), and liquid slugs in the process (unlikely if the knockouts drums are working - triple independent instrumentation on liquid level).
We will assume confidence in the culture on site to monitor the valves and change the oil. But we will assume lower confidence in the vibration and physical measurement tasks preventing a catastrophic failure (they are untested systems). Therefore let's assume a catastrophic failure rate of once or twice/100 years.

If the compressor fails catastrophically – assume a 20-day outage (if the parts are in hand) for the replacement of parts and an additional service cost of 25,000 to build up and machine the bores from the catastrophic failure.

If the work is planned to replace the crankshaft assume an 8-day outage to change parts (if the parts are in hand).

If the work is planned to replace the valves assume a 4-day outage to change parts (if the parts are in hand).

If the compressor is stopped some raw materials are flared at cost of 15,000/day (part raw material loss and part flare gas costs).

As a side note:
We have two sets of valves – one is in use and the other is rebuilt. I would assume if the valve work was only required on a time basis we would wait to the last minute to rebuild the spare set of valves keeping our money available for other opportunities. But as the valves have a probability of failure (not dependent on time) between age replacements we must rebuild shortly after their removal from the compressor.

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

To those reading or adding to the discussion:

To recheck my understanding of the discussion my interpretation is the idea of separating the need to stock parts based on the unexpected or planned replacement. Parts that are replaced on a time basis should be ordered JIT and parts that are "unexpectedly" required (lead time > forecast time) should be in stock.

I have worked in the aluminium industry where I was responsible for over 500+ mobile equipment. An RCM pilot project was completed on one of the fleets of equipment. Originally if total quantities of parts changed/year were categorized 10% were time based (oil and filters) and 90% were on condition. After the RCM analysis we found 50% of the replacements were time-based and 50% on condition. All Time based part replacements were ordered JIT for the PMs. The accuracy of lead times, and PM scheduling (administration) usually had us start "kitting" the PM packages up to 40 days prior to the PM. 40 days may seem like a lot, but the sustaining processes were in their infancy and being developed. And of course to stay in step with the discussion the same parts were held in stock but they were kept for "breakdowns". The role of ordering the parts JIT fell to the Planner and the capabilities of the CMMS – I hate to say that I do not know the details of how this was accomplished (if they ordered the part packages for the PM, if this was automatic, or other) – as this is probably a key point for the discussion.

I should note on the RCM fleet: equipment breakdowns were halved and maintenance costs dropped 20% (nothing too amazing, but an improvement). Additionally as the company was going through headcount reduction (via retirements) there was actually less labour required to maintain the fleet. Maintenance costs: +20% because of part cost increases, - 40% because of labour decreases, NET - 20%

As I am now working in an fixed plant (oil related industry) I find there is greater opportunity for condition based monitoring, but I find the fixed plant culture aids in this – it seems mobile equipment does not get the same respect as fixed plant equipment, and I find spare mobile equipment only exasperates the problem. I believe a fixed plant will have more parts changed on condition and the lead-times and confidence in condition based monitoring tasks will set the stage for how parts are stocked or ordered for replacement. I suggest to those reading this discussion that it is important to decide where to start in the bigger picture before jumping into analyzing which parts to stock and which to order JIT. I suggest you first analyze the equipment for the consequence of failure, develop your Equipment Maintenance Strategies, and identify those tasks that are not bullet proof (but are the best you can do today) and if necessary have some spare parts for those of high risk.

My suggestion actually weakens my compressor example as one can see the consequences do not result in an entire plant shutdown.
So why did I share this example? Because the compressor was recently analyzed with RCM and because how the compressor fails is a good example as to how to it may be necessary to stock parts for condition based actions that are also completed on a time basis.

Why was this compressor analyzed with RCM as again the consequence of failure is not a plant shutdown? Because someone wanted to redesign the compressor when the company lost $800,000 from a failure two years ago, so we did the RCM analysis instead and it told us the compressor is fine, but our Equipment Maintenance Strategies were not.

I like making side notes, so here is another:
We recently started condition based monitoring on the valves of this compressor and not only did we just start the monitoring we also planned the valve replacements due to very high temperature readings. My company saved $700,000 by performing this task. But I like to point out it is very difficult setting the limits of when the work should be completed – perhaps we could have operated another 200 to 5,000 hours one has to balance these decisions against the risks.

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

Thank you Jaz for taking the time and share your experience and points of view with us. Thank you also for the numbers.

Answering your two questions and asking again:

 Unit Purchase Cost refers to the cost of the part (crankshaft or other part). How much is it?
 Lost production cost refers to the loss of one hour production (mean contribution margin of the product mix (mean unit selling price – mean variable costs) multiplied by the mean production hour rate), if the production not accomplished during the compressor stoppage cannot be postponed to some other time such as weekends. Otherwise it will only cost extra time. So, how much does your company loose for every hour the production is stopped due to a failure of the compressor?

I will wait for your answers to these two questions. Meanwhile, I will study your problem carefully in order to give you my perspective.

Rui

Thanks for the comments Rui Assis

I hope this is helping.

I want to make one comment: I think there are two failures mixed into my posts above. One failure is valve related. The other is Crankshaft related. A valve failure can lead to a crankshaft failure.

Valve Failure:
Once every five years caught before going catastrophic
P-F interval (temp): 5 to 40 days
Monitored: Daily
Once every 500 years goes catastrophic – i.e. causes crankshaft failure too.
Age replace every 1.5 years
Planned Outage: 4 days (If parts in hand)
Unplanned Catastrophic Outage: 20 days (If parts in hand), bottleneck is build up machining of bore @ $25,000

Crankshaft Failure:
Condition based replaced on measurement every 20 to 25 years.
P-F interval (wear): 24 to 36 months
Monitored: 18 Months
Fails catastrophically once or twice in 100 years (includes risk associated with a once every 500 yr catastrophic valve failure)
Planned Outage: 8 days (If parts in hand)
Unplanned Catastrophic Outage: 20 days (If parts in hand), bottleneck is build up machining of bore @ $25,000

Lead-time on parts valve or crankshaft: 14 to 21 days (+ two weeks during Manufacturer Shutdown during Summer).

Valve Parts: 15,000
Crankshaft Parts: 65,000
Lost Production costs: $100/hr multiplied by 24hrs/day multiplied by (X + 7 days). Any outage X days long + 7 days to return to normal operation.
Outage Costs: Flaring $833/hr. multiplied by 24 hrs/day multiplied by X days. Where X is the number of days of the outage.

I have repeated some data and clarified others – the remainder of the data is in my posts above.

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

Wow, take a few days off and you guys get down to it!

Rui, last week you asked the following:

quote:

Phill,

I think that I have answered your question. Thanks for your view point on economics and the link, which I will visit with curiousity and interest. Would you like to comment the few lines that I dropped here?

Rui, I hope that you enjoyed the free articles available at the link. My response to your comments (and to a large degree the other posts in the past week) is that there is no doubt that the calculation approach will give you a scientfically supportable answer. There are two issues around this hwoever.

Firstly, is it a better answer? Secondly what question are you trying to answer in the first place?

Let me deal with those in order. At Universoty they taught me all that good stuff about spare parts statistical analysis and so on and even today I will sometimes revert to apply parts of it. But the catch is, in my experience, a 'good enough' answer can determined with some simple back of the envelope type review.

Taking the crankshaft example above, the issue is do we hold the crankshaft and parts or not. There are only two possible outcomes hold or don't hold (I assume that we are not considering holding multiple sets) So, what would tip this decision in favor of holding? Typcially lead time of supply vs reqirement for the compressor. Again typically the cost of the part is far outwieghed by the lost production. So, we were told that the lead time may be 2 -3 weeks. If that is the case then no one will thank you for stopping some part of production for that period (especially at a cost of 15,000 per day for gas flaring). My focus would be on getting that leadtime down to something more workable say overnight or 1 -2 days. Work with the vendor to achieve this. If the issues is delays during their summer shut down work with the vendor the overcoem this (how about consignment stock for two weeks)

My second question was: what question are you trying to answerin the first place? Again, my experience is that there are two outcomes being sought - availability and working capital reduction. The problem is that most people see these as being mutually exclusive but they are not. It is possible to achieve both but you can't do that using the calculation approach that is so popular with the so called 'optimization'.

Let me direct you to two resoucres. In the May/June edition of Reliability Magazine there is an article 'When Optimization Doesn't Optimize'. Thie explains the issue more fully. If you can't get the magazine (and you should subscribe) I will post the article on my website - just let me know if you need that.

Next, download the paper '5 Myths of Inventory Reduction' from www.InitiateAction.com/5_myths.htm This explodes some of teh truisms about inventory management.

My final comment. A lot of what I say is seen as heresy to many that like the 'caclulation' approach to inevntory management. But like all problem solving inventory management is about understanding the issues that tip you from one deciosn to another and the constraints that impact that decision. You don't always need to work from first principles.

Hope some of this has helped.

Hi Jaz,

Please see the attached document with my solution to the valves related case.

I found your explanation a little bit confusing. Nevertheless, I tried to reach a conclusion after assuming a few numbers. The opportunity cost resulting from a failure seems to be too low.

If you find the data that I used incorrect or you wish to address some other case, please change values in the expressions or, if you prefer, just let me know and provide me with the right numbers and I will be glad to perform the calculations myself.

Just one more thing: Calculations using scientific methods are, of course, only pertinent after you have applied RCM analysis to the equipment and reached the conclusion that "it is technically feasible and worth doing" as per John Mowbray's words. It is up to you to evaluate the opportunities to do it. In this case, I pretended this was true.

Regards,

Rui

Valve_failure.zip (30 KB, 14 downloads)

Rui

Thanks for the trial. To clarify the failure costs for the valves.

When the compressor is stopped:

~1000/hr

Plus

~ 17,000 per stoppage of the compressor (process effects that are felt after the compressor is re-started, regardless if the compressor is shutdown 4 days or 20 days)


To clarify the downtime:

4 days if valves are in stock

17 to 24 days if valves are ordered

Jaz,

Now, with these new data, the picture is rather different (valves must be undoubtely in inventory).

Please see the attached document.

The method can be easily extended to other cases.

Regards,

Rui

Valve_failure_1.zip (31 KB, 24 downloads)

I am sorry but I just noticed that I forgot the $17,000 per stoppage of the compressor in the case where spares exist in inventory.

Please change:

Recover cost to 4 days x 24 hours/day x $1,000/hour + $17,000 = $113,000, instead of $96,000 and the final cost of "no stock strategy" to $70,649 instead of $60,957.

The conclusion still holds.

Rui