I would like to share my point of view on why industry is reactive in nature. First let me define what reactive means in simple terms, it simply means addressing a failure when it happens, failure will come first and maintenance will react upon it. Even with the best structure on Preventive Maintenance and top of the line softwares at hand, PM cant capture all the failures, am I right ? Failures are inevitable and occur in three patterns, Infant Mortality, Random Failures and Wear or age-related failures, these were expanded by Nowlan and heap and found out that 6 failure patterns exists. Understanding this failure patterns will definitely allow us to have a better perspective about how to derive a wholistic structure that will capture majority of these failures. PM overhauls and replacement will only capture age-related component failures and likewise induce infant mortality failures but the problem is these only relate to around 20% of equipment failures. Most specially when equipment's are automated and full of electronic / electrical boards and components such as semiconductor, most failures encountered will be random in nature and this is the biggest percentage of equipment failure. This means that Preventive Maintenance cannot capture random failures and maintenance must divert to other options such as predictive, run to fail or modification. A good maintenance strategy will be developing a structured maintenance approach that will more or less capture not only the wear out or age related failures but as well as random and infant mortality failures.

Rolly,

Good to meet you on this forum. I agree with your observations and wish to add the following comments.

1. Our actions must always be driven by what the business needs; to keep society happy, the cash till ringing, and keep them doing so for as long as possible. Interpret these as Technical Integrity, Availability and Asset Life, all of which are driven by Reliability).
2. The timing of most failures is unpredictable using some statistical or other theoretical way (as shown by Nowlan & Heap and others following their lead)). The only way to predict such failures is to wait for onset of measurable failure. We call this strategy CBM or PdM and it is one of the most effective methods. This is a 'reactive' approach, since the 'prediction' is of time of failure, once the (measurable) failure has already started.
3. Some failures have no significant consequence, i.e. they do not matter. In these cases a run-to-failure or breakdowm strategy is a sound business decision. This is a 'reactive' approach.
4. Since less than 20% of failures are related to age-in-service, an age-based approach has applicability in only a small propostion of failure modes. Doing 'pro-active' PMs does not make business sense in the remaining cases.

There is nothing wrong with a reactive approach to maintenance strategy as long as it is applied with careful thought. It is not a 'bad' thing to do. Equally, a 'proactive' PM is not aleways a 'good' thing to do; each failure mode should be treated on merit, not by some mantra.

Reactive strategies can be very cost-effective. Imagine checking the heath of every person in your country every year or month or week to find out whether they are sick. It is usually more reasonable to wait till they show some symptoms of illness e.g., high temperature, low pulse or giddiness, coated tongues or just feeling lousy. We also do not replace the kidney or liver of a person based on his/her age on a proactive basis.

Proactive strategies are necessary when we are trying to detect 'hidden' failures. Taking the human model again, breast cancer screening of women after they are 40 or 50 is quite sensible and cost-effective. That does not mean that some women do not get breast cancer when they are 20 or 30, but their numbers are quite small.

In sum, proactive is not always the right thing nor is reactive always the wrong approach. Each situation has to be weighed on its merit.

Being reactive is the natural course of human behaviour.

Being proactive will encounter hurdles which need to be overcome.

How often is just repair compared to repair and improve or prevent?

Josh,

quote:

Being proactive will encounter hurdles which need to be overcome.

I have no problem with designing out failures in a proactive manner when it makes good business sense; I support that approach.
That was NOT what my earlier note was about. Following Rolly's thesis, which I support, I was talking about the notion that some of us may have that PM is good and breakdowns are bad. Sometimes that is true, but not always.
Where breakdowns have negligible consequences, it is neither desirable to do PMs nor is it justifiable to design out the failure.
Risk reduction must be the driver of our actions not some dogma about proactive approaches.

Josh,

There is no contradictory as to what you say, in fact this is the main purpose of this forum and every forum on reliability, it is easier to repair than to perform a repair and analysis attitude on the part that fail. As my good friend Bob Nelms states that sometimes, we need to take things slowly in order to speed up the process. I have found a common thread on indurstries whether manufacturing, mining, shipping, airline, powerplants, automotive etc., that being reactive have something to do with the culture of the maintenance and operations structure. My point and Vee is that it is much more important to realize the consequences of the failure rather that preventing or eliminating the failure itself.
Most maintenance managers definitely would rather want their people to be efficient at repair rather than having an attitude of analyzing failures which will take time. I would like to correct what I have said before with my fellow rooticians that we are on a loosing edge, but rather there is still hope that it is still possible to let industries understand why they are reactive and improve their approach by understanding how and why failure does occur.

I agree with the idea of RCA but it's not easy to do eg for equipment failures. How can we expedite the process of RCA for equipment failures such as pump mechanical seal leaks? There are so many variables to be considered and eliminated before a RCA can be completed.

My view is that RCA can be done easily for investigation of work process breakdowns but it's quite difficult for investigation of failure analyses which would become RCFA. Consultants who have done RCFA for similar failures or laboratory examination by an expert metallurgist may be needed to complete an RCFA. What do you guys think?

Vee, I agree totally with your post regarding purposeful breakdown maintenance strategy. In fact this is recognized weel in our MMG (Maintenance Management Guide).

Josh,

quote:

it's not easy to do eg for equipment failures.

Analysing physical failures including the understanding of the metallurgy or chemistry is the easy bit of RCA. So I am somewhat surprised by your statement quoted above.
The diffiult part is to suss out the human and latent failures. These are difficult whether you deal with mech seal failures or business process failures or when doing an accident investigation.
If you read Lord Cullens report on Piper Alpha, for example, you will see that a lot of detailed wokr including model testing had to be done to understand the sequence of physical failure. That is similar to the metallurgical analysis of a mech seal. I would argue that in the Piper case, this was the easy part. Understanding why the OIM refused to shut down Piper completely after the first explosion or why the designers put in local start controls rather than in the Control room was in my view, more difficult. Similarly, why did the Dept of Transport inspector not recognize the apalling condition of the sprinkler nozzles and deluge system just a few months before the disaster when many internal tests had shown them to be in very poor condition BEFORE the DoT inspection?. The Cullen recommendation sremoved several common conflicts of interest that prevailed then in the offshore O&G industry.
Back to your concern; if you go back to your query about submersible pump failures a few months ago, you wanted a quick answer. I suggested you need a thorough RCA at site, as the failure could happen again. You did not tell us how you solved the problem or whether you did so at all, so we are in the dark. But I am confident a good RCA would have solved that problem for good.
If Rolly12 is the person I think it is, then as a knowledgeable RCFA facilitator/trainer in your region he may be able to help you. Then you can see the difference it makes.

I haven't got a copy of the Piper Alpha report. Have you got one?

I meant is that there are many factors to be considered for causing the physical damage. For the mech seal leaks, some causes suggested are: cooling water cut off, pump operated against a closed drain valve, motor overloaded, journal bearing failure, thrust bearing excessive play, etc. Another point how to determine which part failed first?

Josh,

quote:

I haven't got a copy of the Piper Alpha report. Have you got one?

Your company library must have copies; else you can buy it from the UK Dept of Energy, ASIN 01 01 13 102X; there is also a brief description on pg 128 of my book.

quote:

I meant is that there are many factors to be considered for causing the physical damage. For the mech seal leaks, some causes suggested are: cooling water cut off, pump operated against a closed drain valve, motor overloaded, journal bearing failure, thrust bearing excessive play, etc. Another point how to determine which part failed first?

Of course there are, and each case, as you rightly say, can be different. That is why you need to analyse each case on its merit, systematically. That is how RCA works finding evidence to support or refute your theory of causual links. You cannot pick up the cause from somebody else's knowledge or experience, though that will provide some theories. But you still have to go through the evidence gathering and analysis process.
At the risk of repetition let me say that understanding the physical chain of events is the easy part.

What about condition based maintenance. Isn't that supposed to be able to identify random failures?

Condition Based Maintenance is supposed to discover incipient failures, giving room for preparing counter measures and interfere in the operation (shutdown and repair) on the most appropriate moment, taking in consideration all actors (spares, delivery time, special services, back-up, manpower resouces, production planning and market demands)

Vee, clarification: What do you mean by the physical chain of events eg for a mechanical seal leak? Pls give examples. Are there several common physical chains of event for a mechanical seal leak?

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

Josh,

quote:

Are there several common physical chains of event for a mechanical seal leak?

A few examples of physical chains:
1. Original fitting incorrect, resulting in incorrect static seal pressure.
2. Misalignmnent of motor-pump set, causing excessive vibration and shaft orbiting.
3. Seal flush inadequqte - incorrect pressure or flow or seal flush fluid.
4. Frequent starts (e,g. by not applying duty/standby), causing start-up wear.
5. Low NPSH causing vapour lock at seals: this causes chattering and impact damage.
6. Incorrect seal design; causes - materials of construction, level of pressure balancing, type of seal - bellows or spring liaded; static O rings or dynamic O rings etc.
7. Seal cooling inadequate, due to coolant temperature or quantity; itself perhaps caused by dirty cooler.
8. Steam quench port out of axial alignment. This will tend to separate faces.
9. Damaged bearing, causing misalignment and vibration, see item 1 above.
etc., etc. ....
As you can see, each item can be drilled down in a tree form to many more 'roots'

Those are cause and effect relationships. There appear to be too many of them to be investigated and eliminated. Anyway to zoom in to the most likely ones?

Josh,
You asked for examples of physical failure chains of events, which is what I have attempted to answer.

quote:

There appear to be too many of them

If you ask for a generic set, there will always be many that dont apply in your situation. You do have to investigate each case on its merit, understanding your local situation.

Hans,


Random failures that provide potential failure can be detected with Condition-Based Monitoring equipment's. To complete the process, there are 3 options to deal with random failures. Run to fail if the consequences of the failure is minimal, Predictive Maintenance for random failures that provide potential failures and Modification when the consequences of failure will not be acceptable at all in order to reduce the risk of failure.

To catch random failure is like taking a polariod camera and take a picture of a lightning. When failures are random in nature then this is when Preventive Maintenance will be at its weakest and industry must understand that adding more activities on Preventive Maintenance will not solve the problem for random cases. I just do not know if people will agree that placing more activities in Preventive Maintenance will likewise induce the chances of Infant Mortality failure therefore :

MORE PM = MORE PROBLEM
LESS PM = LESS PROBLEM

Rolly12

The purpose of condition-based maintenance, when correctly selected, is to perform a non-intrusive set of tests on equipment to avoid PM's that can, and do, generate maintenance-based failures (I do admit, not all CBM is selected that are non-intrusive).

The term 'random failure' does not mean that it is not a detectable failure. It just means that some event may occur that will cause equipment to fail. There are, most likely, indicators of that failure in advance. The proper selection of CBM, where it is appropriate and effective, will reduce the risk of the impact of machine failure.

Howard

Rolly,

Tying in with Vee's original reply on this subject.

IN the attached article I have differentiated between "reactive" unepected and unanticipated failure; and "corrective" failures that are repaired on a run-to-failure basis, or after a predictive or detective task has found a problem. (Or the warning signs of a problem)

There is also a reference in there to the RCM Scorecard published in my book.

I hope it is of use to you.

Letter-Captured_by_Data.pdf (311 Kb, 45 downloads)

Somewhat Commercial

quote:

quote:
it's not easy to do eg for equipment failures.

quote:
Analysing physical failures including the understanding of the metallurgy or chemistry is the easy bit of RCA. So I am somewhat surprised by your statement quoted above.
The diffiult part is to suss out the human and latent failures. These are difficult whether you deal with mech seal failures or business process failures or when doing an accident investigation.

Based on surveying we've done organizations can often get started with the physical problems and develop answers to them. The human and latent failures are more difficult. However, in the end many organizations don't have a good overall process for doing root cause analysis. We're convinced enough of this that we're running a conference/learning event on this subject.

http://www.reliability-magazine.com/