Just what is the best way to address an Infant Mortality Failure, can this be eliminated completely or reduced to a manageable extent.

And what are the major causes of it, I know for a fact that one cause that leads to infant mortality failures are intrusive maintenance.

From the book of John Moubray page 143 states that " It is also borne out by machine operator who says that everytime maintenance works on it over the weekend it takes up to wednesday to get it going again ?

Cheers,

Rolly Angeles

Rolly,

You are correct that intrusive maintenance can introduce infant mortality.
Infant mortalities can also be caused by defects designed into a product. Incorrect installation of equipment (such as misalignment) can also introduce an infant mortality failure.

So what is the best way to address infant mortality, well eliminating the defect completely or managing the failure as you mentioned. Firstly, I would suggest you need to know which failures you have experienced is actually infant mortality by carrying out some Weibull analysis. Once the failures have been identified, what is the consequence or impact of this failure, can we live with it or do we need to eliminate or manage it. Then I would suggest carrying out some Root Cause Analysis on these failures to look for solutions for defect elimination. It may result in redesign of the equipment, introducing redundancy levels, purchasing some superior components, rewriting installation procedures etc……

If you require more information on any of the above please have a look at www.reliability.com.au


Cheers - Gary

quote:

Firstly, I would suggest you need to know which failures you have experienced is actually infant mortality by carrying out some Weibull analysis.

Gary.... I am not a Weibull fan - in fact I think people waste a lot of time and money using Weibull for problems that are inherently simple to solve. Can you tell the forum how Weibull tells you which failures you have that are causing you infant mortality. From my understanding, Weibull will estimate a number of parameters regarding failure data but it will not tell you what mechanisms are infantile.... if that data is available you can determine that before you run the Weibull... hence there is no reason to run a Weibull.
My advice would be to take a look at the data, plot it on a histogram and work it out from there.
Keep it simply Rolly.
Rgds
Steve

I would like to reinstate my question, Is it possible to completely eliminate Infant Mortality Failures or just reduct them to some degree? Is it a wise decision to reduce the amount of overhauls and Preventive Maintenance to reduct Infant Mortality Failures ?

Sorry for the diversion Rolly... just that the Weibull solution is over rated (in my opinion)and I feel the need to speak up and explain simple alternatives.
I have been looking at failure data from all sorts of industries for over 25 years. From this experience, I know it is possible to reduce infant mortality to a negligible level where it appears to be eliminated. I know plants that have done this. They pride themselves on precision maintenance and go to great lengths to achieve zero failure rates. You have no doubt seen clean rooms for electronics and hydraulics... some companies have test programs for components before they install them and the list of actions continue.
For most industrial plants, there is little attempt to achieve reductions in infant mortality - in my opinion again.
Where shut downs are frequent (less than monthly) and infant mortality is a problem (more than two failures in the first day) we strongly recommend a policy / vision / mission of "good day the first day". What does this mean? ... it means some communication of the initiative.. and it means that the failures on the first day become a kpi for a while. Then the infant mortality failures are investigated just like lost time injuries and procedures put in place to reduce them... now not all plants need to go to the extent of total elimination but most times we find that chronic infant mortality problems are cultural and systemic meaning the 1) labour used in the shutdown does not care and management allows this culture to evolve and 2) the procedures to execute and manage the shutdown work are poor.
When I get some more time, I will post some power point presentation slides showing an example of a simple KPI system.
Regards
Steve

The good old "bathtub curve"!
35 years in Refinery Maintenance and yes, you can "reduce" infant mortality but it takes time and effort.
Steve has already mentioned what it takes and the mjaority, I found, is down to good maintenance and workshop practices. Of course operators must also take some responsibility.
In my last refinery, when I first arrived they had an MTBF in their 1400 pump population of just 9 months, the majority of the failures causing this low figure were infant mortality. When I left five years later their MTBF was 36 months. How?
A close look at the workshop practices very quickly revealed one of the root causes. They were "drifting" new bearings onto pump and motor shafts. This practice, to me, is criminal and they only people to benefit is the bearing manufacturer. I bought oil-bath and induction bearing heaters, some good heat proof gloves and completly eliminated premature bearing failures.
Look at your maintenance practices! Not just bearing fitting techniques although this does tend to be one of the main contributors.
Look at alignment. How do you do it? Laser?
Look at all your couplings. What type are they?
Look at your foundations/baseplates.
Look at your piping systems and supports. Are the pumps acting as pipe supports?
Good luck

cheddarcaveman@yahoo.co.uk

Hi Rolly,

I have found over the years that you cannot 100% kill infant mortality. You can reduce the rate with good mechanics following good practices. But even if all your mechanics are top notch and all your practices are "world class" there is always that one golden BB of a manufacturing defect that seems to work it's way in.

Look at what caused your failure. If it was a manufacturing defect ask your parts vendor to help you. If it was caused by workmanship, ask yourself what can we change (that makes sense) to stop this failure in the future.

Aubrey

Rolly,

I believe 2 factors can drastically impact infant mortality, ie. design and installation/commissioning of an equipment/system.
If an equipment/system has adequate reliability (FMEA & thru'put buffer)built into the design, this can reduce infant mortality significantly. The next crtical phase of course, is ensuring best practices during installation and commissioning are engaged to ensure reliable startup and steady state operation with trained personnel.
Reducing infant mortality is a possibility but, weeding it out completely would be a tough shot.

Regards...Rajan

Rolly, I have to agree with Steve about using statistics. I believe that statistics has a place for fine tuning but not for fundamentals. You must aim to reduce the infant mortality rates to a 'managable' extent. You will not have wide enough controls to reduce to zero. Every step in the process of performing a repair is subject to human error. Note, however, that it also applies to the spare parts used. The manufacturer has the same issues with producing usable parts. I often have to remachine parts, or ask for replacements if remachining is not an option.
Everyone who is involved in maintenance is subject to errors which can have an impact in the repair. If the levels of competency in the chain of involvement is adequate or high, the incidence of infant mortality will inevitably reduce.
Regards,
Joe Mc Cormack

Folks,
Maintenance or operating errors are NOT the only reasons for infant mortality. Mechanical components that slide with mating parts align and adjust themselves, selectively wearing out 'high spots', a process known to all of us as 'bedding in'. During the bedding-in process, reliability improves continuously and finally settles down at some value.
In addition to this baseline level, there are additional sources of failure, relating to the quality of maintenance, referred to by Rolly (Moubray, page 143), AND quality of operations, a major source in itself.
The quality issues can be minimized by not doing unnecessary maintenance. Economics will drive the quality of manufacture of the parts and their initial assembly, so the baseline 'bedding-in' issues can never be adequately solved. Unnecessary part replacement can cause bedding-in problems, and field work can create misalignment, incorrect fitting, dirt ingress and bad shutdown/startup.

Doing more maintenance is commonly seen as doing the right thing - I would argue that the right level, quality and frequency is what really matters. Leaving a well running machine alone is not a bad thing, but dont forget to check and confirm regularly that it is in fact running well.

Hello Steve,

Apologies for the delay in my reply, I have had no web access for the last week.
Weibull analysis is a relatively simple methodology done at the failure mode level. It’s important to know how the equipment is failing before you can determine what strategy to implement. How can you tell from Weibull which failures are causing infant mortality? You simply look at the results from your Weibull plot and take note of the beta value. If you have a beta of less than 1, this is infant mortality, if it is equal to 1, this is a random failure, if it’s greater than 1 then you have a wear out. Once you determine which part of the bathtub curve you sit, you can now determine which strategy you can adopt.
If you are going to look at the data, plot it on a histogram and work it out from there, why not use the same data to quickly run a Weibull plot and get some quick feedback on what type of failure you are experiencing.
For more information, I would recommend reading “The New Weibull Handbook - by Dr. Robert B. Abernethy more details at http://www.barringer1.com/tnwhb.htm

Cheers Gary

quote:

Weibull analysis is a relatively simple methodology done at the failure mode level.

Thanks Gary... based on what reference point are you saying that Weibull is relatively simple ? Perhaps it is simple relative to calculus of vector functions which I did in 4th year of my degree. Not so simple for Rolly who may not have been trained in Weibull.
But Rolly(or his client) "knows" he has an infant mortality problem because he feels the heat from production after every shut... so you are going to tell him to run a Weibull plot just in case the high rate of failures he sees in the first few days after every shut are a statistical abberration.

quote:

If you are going to look at the data, plot it on a histogram and work it out from there, why not use the same data to quickly run a Weibull plot and get some quick feedback on what type of failure you are experiencing.

Because a histogram can tell you more information that a Weibull plot... for example, a Weibull plot can only show you patterns related to two or sometimes three failure mechanisms... a histogram can show you many more. And with a histogram, you are more likely to see where the data is incorrect, and the accountant can understand the histogram and so can the mechanics and the operators.
To me, running a Weibull plot may make the engineer feel smart but that is about all in this case... just a waste of time in my opinion.
The data is staring Rolly in the face - all he has to do is diagnose it (the easy bit) and start programs of elimination (the hard bit).

Regards
Steve

Steve,
We dont need advanced calculus or vector algebra or other fancy maths to do a Weibull analysis. All we need is a log-log Weibull chart paper, sequential run lengths, and premature repalacement during PMs or early retirement information. All this should be in the CMMS.
We can do away with the special chart paper if we have Weibull software.
It has its place in reliability analysis, so I am not sure why you argue so strongly.

quote:

Weibull software.
It has its place in reliability analysis, so I am not sure why you argue so strongly.

Vee - Weibull has a place but nowhere need the prominence that the vendors will have people believe.
This post is a classic.. Rolly is asking about solving infant mortality problems and the first response we get is go complex with Weibull.
Gasp... if Rolly wants to identify the problems and solve them, he needs to ask the right people the right questions and these people are not mathematicians. They are operators and technicians who will not make sense of a log-log Weibull chart paper without some training from an expert.
I think Rolly needs to get some data about what is failing early (infant mortality), and arrange the information into a format the operators and technicians can understand. Then, get some buy in to solve the problem. Then of course put in some project management etc etc.
I have seen papers and presentations conference after conference showing how Weibull solved this problem and solved that problem.... truth is, my histogram would have solved them too... and you may recall my post last year with some real life data... the Weibull analysis came up with the wrong answer... you see in that case, there were three mechanisms of failure and the Weibull guys made the assumption there were two.
Weibull is a method that approximates certain types of failure patterns at a high level and provides some mathematical insight into what may be going on where there are one or two (and at a stretch three) mechanisms of failure. Most of the time we dont have enough decent data so take the significance of the Weibull distribution seriously..part of a different debate.. which I shall keep away from at this point... I much prefer histograms with lots of labels and questions.
But Vee - if you say that, for a lot of engineers, Weibull is fun to do - I will certainly agree with you. And there in lies one of the problems... engineers will buy Weibull because its fun and stimulating, but not buy into a simple excel histogram... so with all the hype and referred value, the juggernaught continues.
At the end of the day I am sure there is room for many opinions.. Vee you asked for mine and you now have some of it.
Thankyou for the opportunity to get on my soap box and throw my two bobs worth in.
Regards
Steve


PS - from Wikipedia.
The term juggernaut is used to describe any literal or metaphorical force regarded as unstoppable that will crush all in its path. In Britain, it is also used to refer to any large and heavy lorry.

The word is derived from the Sanskrit Jagannātha, meaning "Lord of the universe"; it is one of the many names of Krishna from the ancient Vedic scriptures of India. One of the most famous of Indian temples is the Jagannath Temple in Puri, Orissa, which has the Ratha Yatra (chariot procession), an annual procession of chariots carrying the murtis/statues of Jagannâth (Lord Krishna), Subhadra and Baladeva (Krishna's elder brother). During the British colonial era, Christian missionaries promulgated a myth that Hindu devotees of Krishna were lunatic fanatics who threw themselves under the wheels of these chariots in order to attain salvation. Such a description can also be found in the popular fourteenth-century work "The Travels of Sir John Mandeville." The fact is that devotees have sometimes been crushed accidentally in the past as the massive 45 foot tall, multi-ton chariot slipped out of control. Many have also been killed in the resulting stampedes. The sight led the Britons to use the word "Juggernaut" to refer to other instances of unstoppable, crushing forces.

In modern times, the government officers and temple priests managing the festival take elaborate precautions to protect people from injury during these processions.

The Ratha Yatra festival has become a common sight in most major cities of the world since 1968 through the Hare Krishna movement. Its leader A.C. Bhaktivedanta Swami Prabhupada successfully transplanted the festival which now happens on an annual basis in places such as London, Paris, Toronto and New York.

Steve,

I agree with you, in that we should not needlessly complicate things when simple solutions are available. A simple histogram tells a good story, but there is absolutely nothing to beat a chat with the operator or maintainer to understand what is really happening.

Rolly ia an experienced and knowledgeable engineer; I have a great deal of respect for him. I think he was just throwing up a ball to check the reactions. When you do that in a forum, you should not be surprised if you get a range of suggestions. So the Weibull foray is best seen in that light. That is why I was surprised by the strength of your comments. I use your histograms and paretos a lot, but I have also done quite some Weibull work. By the way, we need Weibull parameters any time we do mathematical modeling, so I have a healthy respect for those who work with it.

The wikipedia intrigues me; as one who comes from India, and whose name is a variant of Jagannatha, I know a little bit about the Juggernaut story. While it was entertaining, I did'nt see the context. Perhaps you can explain.

Looks like my plans of getting some important work done before the office opens are not going to work out... impulsive thing this forum...

quote:

By the way, we need Weibull parameters any time we do mathematical modeling, so I have a healthy respect for those who work with it.

Sure you do - but where do you get the data for the parameters from and what is the statistical significance of the outputs based on your inputs ... should Rolly be using modelling tools to solve infant mortality?
On the juggernaut subject... I used the analogy that Weibull analysis is a juggernaugt... "which by Wiki definition is a literal or metaphorical force regarded as unstoppable that will crush all in its path". I hear people all over the world yapping on about weibull this weibull that ... it seems to feed on itself and grow.. people get sucked into the path and become clones of the others..
Sure it has a place but not to solve every problem that is on the pathway.
If I achieve anythng by writing on this forum, I hope to slow down the juggernaught.
Steve

Steve, Vee and all,

With high respect to the late Wallodi Weibull and what his statistics can share in the maintenance arena, yes I know for a fact it can help and as well as your Basic Statistical Tools of SPC called Histogram.

But the point I would like to discuss here is more of a Higher Heirarchy Decision process that helps a lot in contributing to INFANT MORTALITY FAILURES and this is what I am emphasizing dearly in all of my trainings, to let the people I teach understand what infant mortality is and how it affects us and how maintenance and operations become mortal enemies when maintenance will try to get the machine for PM which operations will deny.

Let me be concrete about what I am trying to say and what you can suggest. Let me call this the ADD ON PM SYNDROME, that leads to infant mortality failures, simply stated

MORE PM = MORE PROBLEM
LESS PM = LESS PROBLEM

When a dear customer walks on your plant and notice something on your equipment, the first option here will be to check if this was included in the PM checklist, when a quality audit is in place and found some leak underneath, then the option once again will be to check if it was on the PM checklists and again and again the checklist multiply until there come to a point in time that maintenance cannot perform all the activities because they are outnumbered.

Now these people, customers, quality inspectors, Top Management, or even your CEO, failed to understand that there are actually 6 failure patterns and every single component found on your equipment will behave according to the 6 failure patterns. In my case I think these people are the main source of Infant Mortality Failures since they just do not accept the fact that these type of failure exist.

I recall performing a classical SAE JA1011 RCM on a piece of equipment which has around 350 maintenance tasks to be performed for the year, and we have reengineered it and dramatically reduced the tasks by half. However, the RCM analysis was never implemented, and they were afraid to implement it, why because most of the maintenance task that are listed was provided by their customers during a customer audit and removing this type of activity means deviating from their customer demands.

Steve, I always believe that your weibull will identify the infant mortality parts, but people will likewise do nothing about it since this activity was place by a KING who simple does not understand INFANT MORTALITY FAILURES, and the rule of law will always be to OBEY the KING.

I hope you see where I am coming from, what happens everytime a machine completed its PM run, and why start-up failures are just a normal trend as well as debugging and so on. The problem never seems to go away since they do not understand infant mortality at all most specially the KING.

Regards,

Rolly Angeles

This message has been edited. Last edited by: Rolly12,

Rolly, your RCM results are not implemented due to worries etc?

This seems to indicate lack of leadership and willingness to improve things.

Why can't they discuss the rcm results with the customers and maybe the product price can be reduced or so?

I don't quite understand why a customer dictates the PM program? A customer audit like the ISO9000 one may ask whether you have a PM and the avalability and reliability (the end result/bottom line) but not the details on how to do it.

Another thought, faced with this scenario, why not bring the matter up to the management committee or board of directors for delibaration if the Maintenance Manager cannot decide on himself? This will make the decision making based on collective teamwork.

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

We have to obey the King? If we always have to obey the King, there will be no revolutions and no republics.

This seems to suggest a mindset problem of not challenging the status quo. The organization in question may need more than understanding the meaning of infant mortalities. The maintenance people may understand the idea but how about the rest eg the CEO?

If the maintenance dept implements the RCM results without agreement by the rest and something wrong with the machine, most likely the Maintenance manager will be pointed to as the culprit (which I think he tries to avoid).

I think the organization needs more than just a maintenance training.

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