Gentlemen.

There seem to a a tendency by reliablity and maintenance practitioners to overstate the actual availablity of plant we run.

When we try to reconsile the plant availability with actual throughput there is a big discrepancy that can not be explained by merely accusing production or operations of inefficiency.

How can we calculate plant availability in such a way that we can relate the throughput achieved by de-rating plant availabilty by some reasonable co-efficient [typically in the 90 percentile] to come up with actual production throughput.

Ecky

Hi Ecky,
I think it depends on the management - some report acurately. The only way to get the right figures are to reconcile with production output where this is possible. In manufactureing for example, you need to get the rate and the time and the amount produced. if the rate and the time dont multiply to the amount produced, then you can work out the losses.
I think that ego's and KPI's often drive people to misrepresent the facts.
Rgds
Steve

Ecky,

Do you mean the plant does not produce as per the nameplate capacity and thus you to de-rate the plant capacity?

quote:

Do you mean the plant does not produce as per the nameplate capacity and thus you to de-rate the plant capacity?

No, Josh.

What I meant was that
If we de-rate the plant availability by some factor, usually less than unity, this should cater for the inefficiencies occassioned by operational personnel.

When the derated plant availability is multiplied by the rate at which our plant should work, this figure should agree, within close limits, with what was actually produced.

I was led down that stream of thought when I noticed that one company was consitently returning plant availibities in the upper 90's. On comparison with what was actually produced, the plant was only effectively using at best, 60% of its installed capacity.

The discrepancy was too wide to be explained merely as operational ineffiencies. Thus I suspected that the mathematical model used to come up with the plant availability was too pie in the sky and of no use as a management indicator.

quote:

I think that ego's and KPI's often drive people to misrepresent the facts.

I cant agree more with you Steve.

Maybe somebody out there has created a realistic model for calculating the plant availability, which he is willing to have critiqued by others

Ecky

It seems that you are comparing plant availability with plant capacity which are two differrent things.

How do you figures for each? What type of plant is this?

Plant availability is basically uptime over a year if running all year long.

Plant capacity is the production rate per day or year for example.

quote:

It seems that you are comparing plant availability with plant capacity which are two differrent things

This gets interesting Josh.

quote:

Plant availability is basically uptime over a year if running all year long.

Plant capacity is the production rate per day or year for example.

By way of example. If I am to beleieve my Maintenance Engineer, he tells me that the plant was available 97% of the times during the year and if my plant is capable of producing 100 widgets per year, I can not expect to get out all the 100 widgets, because the plant was only available 97% of the time.

At best, I can expect 97 widgets if no overtime is worked and operations are working at 100% efficiency.

But if I rate my operations department at 90% efficency, then my expected production will be around 87 widgets.

Again, if at the end of the year, I was only able to produce 67 widgets, having accounted for all the other losses, I will suspct that either my production efficiency or my plant availability figures are wrong.

It is the issue of trying to reconcile the plant availability and plant capacity versus what was actually produced on the ground which led me to realise that maybe as Maintenance and Reliability Practitioners, our measure of plant availability is not consistent with reality.


I stand corrected


Ecky

Ecky,

Availability, which usually means 'time availability' will always overstate the volumetric production capacity. This is because
1. Startups and shutdowns are not instantaneous, there is a ramp-up ramp-down time that needs to b factored in. Thus two identical Plants with the same availability figure, say 94% can have different outputs, simply because one has more trips, so more startups are required.
2. In a continuous production Unit, any temporary bottlenecks that appear upstream or downstream (Such as product storage capacity or a trip of a downstream unit) can cause a slowdown. On paper the availability of the Unit in question is still 94% but it is not producing useful product since we cannot process it downstream or store it.
3. If the process is not producing product to the right quality, due e.g., to a process upset, catalyst degradation, furnace heat transfer capacity loss etc., some additional reprocessing will be required. This means a loss of capacity, even of the Unit is working at the same 94% availability.

In the assembly line scenario, the concept of OEE covers all these aspect neatly (see e.g., Robert Hansen's book on OEE - ISBN 0-8311-3138-1, Industrial Press, NY or most books on TPM for details). In a continuous process plant, we use the concept of System Effectiveness (or volumetric capacity availability) in a similar manner.
The factors that affect useful capacity differ between Units and over time. We cannot apply a single universal factor to 'correct' the availability in line with your wishes. We need to understand each cause and take the required actions case by case.

Yes, Ecky, you have to ask for any production slowdowns in the 97% plant availability which may account for the discrepancy. The slowdown is similar to you what refer to as efficiency.

We have a dilemma whereby the OEE in our different factories around the region are compared. This is used as one of the metrics to judge a factories performance (there are obviously others).
Although we are perfoming above average on this count we are at a disadvantage to some of these other plants who run a continuous line on one product (variant).
We on the other hand run multiple variants on one line and because of the frequent necessary changeovers are therefore disadvantaged when it comes to the OEE calculation.
I have long argued there must be some way to introduce a factor for the no. of changeovers that can even out this discrepancy and show how flexible we actually are.
Our advantage is we can achieve very quick cycle times thereby reducing our inventories and stored stock and our machine utilisation is right up there.
The downside is our change parts wearing out!

Mike.

Mike,
I can understand your frustration about your position:

quote:

I have long argued there must be some way to introduce a factor for the no. of changeovers that can even out this discrepancy and show how flexible we actually are.

, but I do not think that is the right solution.
Having many changeovers by design must be a policy decision - it is probably very profitable to operate that way in your context. Your peers may be losing out on the $$$ front by having a single product continuous line, but that is their policy.
I suggest you focus on the improvements you make in OEE every year. You cant change the hand you have been dealt, but surely you can learn to play it better!

quote:

Originally posted by Ecky:
[QUOTE] I will suspct that either my production efficiency or my plant availability figures are wrong.

It is the issue of trying to reconcile the plant availability and plant capacity versus what was actually produced on the ground which led me to realise that maybe as Maintenance and Reliability Practitioners, our measure of plant availability is not consistent with reality.

Ecky

Ecky,
I would agree that there are cases where the availability figures are intentionally or accidently mis-represented. There are several issues that lead to different measures of availability e.g., interval choosen and how downtime is defined.

Availability is always measured over a time interval, i.e. daily, monthly, annually. The longer the interval, the lesser the variability.

I have seen the availability of a continuous operation plant close to 100%. While it was accurate for the previous 3 months, but not accurate for the whole year. Since the plant was back to service after a long shutdown of 20 days.

The downtime put in availability calculation shall include all sources of downtime, including planned outages. Including and excluding planned events can cause significant difference.

Therefore i would suggest you to check the accurateness of avaialbility figures of your plant, and see how they have calculated it.

Regards,

Mohammad

Mohammad,
Please include planned downtime, always, as downtime remains downtime, planned or otherwise. It is a good practice to divide planned downtime by shutdown interval(in days) and multiply it by 365 to get an 'annualized' planned downtime figure.

Excluding planned downtime shifts the focus from duration management. That is an important area to consider. Getting a 100% availability by excluding a 20, 60 or 90-day shutdown does not make business sense.

Vee,
Yes, i agree, thats what i have said, all the sources of downtime shall be included for availability calculation. Since during the downtime, the asset/system was not available for production, therefore that time shall be considered as unavailablity.

There are also some other knock on effect issues like power failure/trips related to transmission and other sources causing the shutdown of asset although the asset/system was available for production. Its important to address these issues for performance measurement. In this case measuring Uptime will be a better idea rather than just measuring availability.
Rgds,
Muhammad

Mohammad
Can you kindly do a re-take on Uptime. I have heard it mentioned a number of times and have a general and intuitive understing of the concept.

But if, as a measure, it can relate the work that we do on the plant in a better manner than plant availability does, then, it is a good measure.

But how do you

• scientifically define it
  
• Calculate it
  
• Use it as an effective engineering management measure

Ecky

Ecky,
Uptime is often confused with availability, and sometimes used almost interchangeably.

Availability is the measure of "ability to produce", whereas Uptime is the measure of actual running time while the asset /system actually produced.

I have seen a lot of cases where production machinery had availability figures >95%, whereas the actual production figures were low. When investigated, it was found that the atual uptime (eg running time) was approx 90%. The asset / system was down due to power constraints or there was spurious trip causing the downtime although the assets were available.

Uptime is actually the percentage of running time and calculated as,

Uptime = (Running time / Total time) * 100

If a machine operated for 22 hrs in 24 hrs, and one hour it was down due to power supply constraint (or any other external reason), then the uptime would be (22/24)*100 = 91.66%

Whereas, Availability would be 95.8%

Availability figures sometimes hide the actual uptime performance and therefore donot show realistic picture.

Hope this answers your question
Rgds
Muhammad

Muhammad,

quote:

In this case measuring Uptime will be a better idea rather than just measuring availability.

Both matter, Uptime for the reasons you state, and Availability because that is the main purpose of Maintenance. This is the 'output' of Maintenance (and Operations), so it needs to be measured.
The gap between Availability and Uptime is also one worth analysis, as it is a source of inefficiency.

Uptime does not relate completely to production outputs; even with high Uptime you can lose production, due to slowdowns, setup time and Quality problems. Thus with frequent but short duration trips, both availability and Uptime may be high, but production can be low due to
- Ramp down and ramp up time at each trip
- Quality problems caused by upsets due to the trips

Bottlenecks, both upstream and downstream can cause slowdowns but not stoppages. Thus we can have high Uptimes and low production.

Thank you gentlemen for for enlightening contributions. But, some issues do arise from the same.

It is common practise to treat a redundant unit as available when the primary working unit fails and we activate the stand-by unit (e.g. a two pump work station, when the active pump fails and we switch on the standby unit). I am of the opinion that such practice is practically wrong, for it paints availability in a better light than should be.

I would propose that people should take the conservative approach in the case of the redundant equipment and treat the unit as not available. Such approach would see the unit that failed being restored to an operational state in the earliest possible time.

The other alternative would be to perform reliability calculations for the redundant unit and assign a probability value based on the number of redundant units and their relationships. This route may not be popular, as it may be complex to the majority of the plant personnel.

Given a reverse situation of a single work unit connected in series with other work stations, where this work station happens to be a bottleneck station, how should we treat it in the event if fails.

If we are to take the worst case scenario and note that this single station stops the entire plant for the duration of the breakdown, then we will have to treat all the machines as down, if we are to extend the logic of the standby unit. This will cause our availability to fall significantly (due to a summation of downtimes)

What I have seen in practise is that we want to isolate this particular unit which suffered the breakdown and capture it as if it was the only one that was not available for production during the duration of the breakdown. This again tends to paint too rosy a picture. I would think that the plant stopper should be reported in that context, i.e. that it caused stoppage to the entire production chain and we report a loss of time for the entire plant, necessitating that we report an outage for the entire plant.

This will cause a depression in our reported plant availabilities, and cause maintenance practitioners to work flat out to restore the plant to an operational status. Maybe, it can cause a better correlation with production

I wonder how other practitioners treat the two scenarios that I heve presented above.

Ecky

The other alternative would be to perform reliability calculations for the redundant unit and assign a probability value based on the number of redundant units and their relationships. This route may not be popular, as it may be complex to the majority of the plant personnel.

Calculating availability and plant capacity does get complicated as many times there are work arounds, standby, surge or parallel paths and voting relationships to consider. We do a lot of complex plant modelling but we use (and sell) AvSim+ software. The benefit of using good software is that it simplifies complex relationships. In AvSim a graphical Reliability Block Diagram is built to represent the plant/process relationships and interdependencies.

Reliability and Maintainability is added to each block including ramp up/down time, max capacity, standby condition, switching delays, start up failures are all added.

The results are availability, reliability, production capability for the whole plant. But you can drill down through each system to the lowest level.

The bottlenecks are easily identified.

We are using AvSim+ at design stage to help designers understand likely production capability of their design- given reliability characteristics of each piece of equipment and options for maintenance. It is far too complex to assess any other way.

We also use it for existing plants/processes to identify bottlenecks and understand the impact of modifying plant or changing maintenance strategies such as shutdown intervals.

Great thread and you have given me a great topic for a webinar!

Thanks

Here's the link to a paper by Mick Drew with detailed information on availability/capacity relationships and modeling.

http://www.reliability.com.au/index.asp?pgid=147

I hope you find it useful.

Cheers Gary

Ecky,
You say:

quote:

It is common practise to treat a redundant unit as available when the primary working unit fails and we activate the stand-by unit (e.g. a two pump work station, when the active pump fails and we switch on the standby unit). I am of the opinion that such practice is practically wrong, for it paints availability in a better light than should be.

Why so? The definition of availability is just that; that the item is ready for use if required or it is already in use. I don't see the need to redefine availability because of a problem that you believe exists. Such a move would not help you with your original problem either, since you still have to account for slowdowns as well as quality related issues.
In my view it is better to get a clear picture if what each term means and use the right one for your purpose. There are enough definitions in various National and International Standards; let us just use them.