FMEA Analysis

I would think the start point would be failure data, from the field or from in house testing.
Use the 80 20 rule to analyse your basic failure data.
E.G Concentrate on the most common failures (Basic problems, not detail, as you havn't woorked out the detailed cause yet).
So, say the most common reported problem is 'motor doesn't run'.
Analyse as many of these as possible to narrow down the cause, then attack the most common of these actual causes. It may be a thermostat, the motor or whatever.
If it's the thermostat, you then break it down further...what part of the thermostat? Why? etc.

The real answer to "Where do I start?" is...
It doesn't matter, the important thing is to actually start....

If you mean the theoretical side of FMEA then you need to sit around with guys who install/use the equipment and look at the most likely possible things which can go wrong. EG, Swapping of connectors/wires, intallation problems, blocking of vents etc.. Or motor failure, thermostat jamming etc. And analyse what will happen if these go wrong.
Don't start with the obscure and unlikely, Alien intervention should be way down on the list....

So to sumarise, i'd start from top down not bottom up.
Del

Del's given a good start to the process. I'd agree - start at the highest level and work down.

3) is there a strong difference between DFMEA and PFMEA?

Absolutely!

DFMEA = Design Failure Modes Effect Analysis

PFMEA = Process Failure Modes Effect Analysis

The TS16949* guidance booklets have reasonably clear explanations on the details of carrying these out, it might be worth buying them, even if you're not going for TS approval.

Simply:

DFMEA is what you use before you design something, to identify the likely causes of failure to do what you want it to do and to allow you to design them out first time round, thus reducing the costly prototype/testing/redesign loops.

PFMEA is what you use to before you start making something, to identify the likely causes of making it wrongly and to design them out of the process before you start buying and installing machines, reducing commissioning/rework cycles. It helps reduce your variability too and provides the basis for your operator training and control plans.

If you're being pendantic (me? never) then you "can't" do a an FMEA on an existing product. However, none of us start from a clean sheet, and you'll learn a lot from doing them on products you already know.

4) how to calculate the critical RNP? In another word how to understand that this or that value of RNP is critical for our product?

The RPN is arbitrary - use it to identify the top 20% (back to the Pareto rule Del mentioned) of risks. Eliminate/reduce those, then work of the highest 20% of those remaining. Repeat until you reach the point of diminishing returns.

5) our product consists of mechanical, electrical and electronic components. How to FMEA evaluate their common job?

If you start from the higest assembly level and think about how that could fail, and what the cause(s) of those failures are, you'll find the different components will populate the causes column and you can then separately DFMEA those that occur most often.

AND FINALLY

Don't do this alone - and don't involve everyone. Groups of 4 are most efficient - make sure you have a range of expertise about the product or process in question.

Set a definite timescale. Now double it. And again. The first ones take time as you adjust to the thinking method. Schedule the time in blocks of 1 - 2 hours over a couple of weeks. When you get bogged down in the process, stop and ask the basic questions again: in what way can this stop working? What causes that? How can we prevent or detect the failure?

You don't have to fill the columns out L to R - in fact, that's pretty much impossible.

Good luck - and enjoy.

*TS16949 = the automotive equivalent of ISO9002; it has all the ISO stuff, plus extra bits for the automotive industry.

Del the cat and English Rose,

Thank you for comments.

I have to continue the discussion.

ER said:

"If you start from the highest assembly level and think about how that could fail, and what the cause(s) of those failures are, you'll find the different components will populate the causes column and you can then separately DFMEA those that occur most often."

You know, I tried to start DFMEA from the highest assembly - our cooling system in whole and I've found that it is the wrong way. I couldn't to describe our system in function except three of them:

1) cool the transported product

2) environmental acceptance

3) safety.

Our system consists of a plenty of mechanical and electronic devices. Where are their functions?

I tried to start a DFMEA analysis from sub-assemblies and felt much better because this way I describes every element. But in this way I had feeling that I've something missed.

What is the right way?

Regards, Vadim

Ultimately, the right way is the one that works for you.

Here's an idea of what we meant (if I can presume tot speak for Del):

Function: Cool the product

Failure Modes:

1. Product too warm
2. Product too cold (i.e. freezes)

Causes of Failure Modes:

Causes of 1:

1. Thermostat Faulty
2. Power interrupted
3. Heat exchanger faulty

Causes of 2:

1. Thermostat faulty
2. Heat exchanger faulty
3. ECU failure

Detection Methods:

<blah>

------------------------------------------------------

Can you see where this is going? You then start DFMEAs from thermostat and heat exchanger (in my example) as they effect both. Look at your scoring for the ones with highest risk factor (RPN) to distinguish the next important.

This avoids you spending loads of time on a component whose failure doesn't dramatically affect the working of the unit and missing one that does.

BTW 2) and 3) in your post are not functions.

Thanks again ER.

I found your idea pretty well.

There is the small problem concerning:

"BTW 2) and 3) in your post are not functions."

In the FMEA FORM Table there is the first column called: "Functions/Requirements" (probably I found the wrong FMEA form).

So, as I see, I have to describe the possible functions and requirements for a system or unit. Doesn't it?

Any way, this analysis should carry out the quality department. Because they are in response for implementation the ISO/TS 16949.

I am just Design Engineer.

Regards, Vadim

Are we talking about FMEA or FTA?

In FMEA we go upwards.

Take a component - for simplicity, take only critical components (column-1)

Coil

define its functionn : (column-2)

to carry the fluid (row-1a)

To cool by conduction. (row-1b)

...

Define the failure modes (here there is a bit misnomer) The failure mode is not how the coil can fail. The failure mode is how the coil will fail in its properties. Ie what are the possible defects that can be created in it.

Material not properly heat treated.

Handling damage during bending.

Kink at the bend radius

Handling damages during assembly

...

Effect (column-4)

How each failure mode (ie defect) will affect its function

After this there are several templates

Say

Column-5 : Severity (Rank 1-10 or 1-5 1= least almost no effect on performance, 5 or 10= makes the component non-functional). Damage to personnel is a part of this

Column-6 : Occurrence - this is more of a brainstorming for new product, available for similar/ existing product 1: rarely occurring, and even when they occur, no effect on the product cycle time/ down time; 5 (or 10) : opposite of above in both aspects.

Col-7: detectability: in case the failure (defect occur) how quickly it can be identified?

1: immediately, in the same workstation

2: after some time, may be next work station

5(0r 10) : will definitely pass on to customer.

Multiplying the three gives you RPN.

Then the plugging starts based on the RPNs: and also theoretical RCA (root cause analysis) - this defect could occur in future due to ... insufficient information, not properly commented on drawing, operator(human error), machine error,....

Sorry the post was a bit long

Are you doing a DESIGN FMEA or a PROCESS FMEA ?? See my first post for the differences.

The things you've listed under "Failure Modes" are in fact failure causes. Modes are what the user sees (or what he tells you when he phones to complain).

At the end of the day, it's a tool to help you design better products and manufacture more consistently and efficiently. If it does that for you, then it's worked. They're not just to frame for the wall (or the auditor).

http://www.qualitytrainingportal.com/resources/fmea/fmea_process.htm

http://www.siliconfareast.com/fmea_quickref.htm

Identify all potential Failure Modes associated with the product or process.

A failure mode is defined as how a system, product, or process is failing.

Now here arises some confusion in the semiconductor industry, which usually measures its failure modes in terms of how the product or process is deviating from its specifications. A product or process can have hundreds of different failure modes based on this definition, most of which are highly correlated to each other because of a common failure mechanism behind them.

A failure mechanism is defined as the physical phenomenon behind the failure mode(s) observed, e.g., die cracking, corrosion, electromigration, etc.

For each of the failure modes in Column 2, a corresponding effect (or effects) must be identified and listed in Column 3 of the FMEA Table. A failure effect is what the customer will experience or perceive once the failure occurs. A customer may either be internal or external, so effects to both must be included. Examples of effects include: inoperability or performance degradation of the product or process, injury to the user, damage to equipment, etc.

http://www.npd-solutions.com/fmea.html

Identify Failure Modes. A failure mode is defined as the manner in which a component, subsystem, system, process, etc. could potentially fail to meet the design intent. Examples of potential failure modes include:

• Corrosion
• Hydrogen embrittlement
• Electrical Short or Open
• Torque Fatigue
• Deformation
• Cracking

http://www.hownwhy.com/FMEA/fmea_a_quick_refresher.htm

1. List what can go wrong at each process step (for Services and Processes), or what can go wrong with the components / sub assemblies
   (for Products). This becomes your failure mode (in a bottoms up approach).
2. List effects for this Failure Mode as they pertain to the performance of the object

I could have mentioned quite a few papers on this from ASQ, but most of those are restricted to registered users only, and hence OP may not have access to them.

This confusion between the failure modes and the effect is prevalent. The failure I have seen during my lectures are always associated with the end product failure and not in the failure to meet the basic requirement.

Modes are what the user sees (or what he tells you when he phones to complain).- these are really the Effect of the failure modes.

There is a FMEA software which will help you step by step of doing the FMEA for the process or design, sorry I forget the name of the software. or alternatively use your knowledge and try it out to see how it work. One thing for sure that you need to get different departments involve in the process eg for PFMEA you need at least quality department, manufacturing engineering, production and design engineer if you can to attend the meeting to work out how to prevent the failures happen during manufacturing process. I would try out with one simple process with simple part first before tackle the complicate one

Cheers

OP,

There is an Excel add-in that may help. View the product at

http://www.sigmazone.com/SnapSheetsXL_officemarketplace.htm

and click on the link that says "FMEA features and screen shots". Don't know if this is what you're look for since it's FMEA and not DFMEA or PFMEA, but maybe it'll give some ideas.

Good luck!

Logan

Vadim,

Check MIL-STD-1629 for starting point.

This should lead you to MIL-HDBK-217, "RELIABILITY PREDICTION OF ELECTRONIC EQUIPMENT".

Note that the process is time consuming. I seem to remember that there is software to setup the process.

Google "FMEA".

Best of Luck,

Alex

As has been said already there is definitely a difference in a DFMEA and PFMEA and that is in a DFMEA you are defining the failure modes in respect to the materials being used in your design whereas in a PFMEA you are are defining the failure modes in respect to what damage is done by each operation in your process as it comes in contact with the part being made.

To give you an example of one many would screw up and I see is still confusing by some of the replys you have recieved already.

I previously worked for a major manufacturer of pressure sensors. At one point in the process a ceramic capacitive subtrate has loctite adhesive applied onto it and then a ceramic hybrid (electronic circuit) is placed on and bonded. Later on in the process after the capacitive leads have been soldered to the hybrid and the sensor package is thermocycled prior to calibration the hybrid could crack if excessive adhesive was applied.

Many engineers would put the failure mode as the cracked hybrid but in reality that is an effect seen by an downstream 'customer' at calibration or end of line test or even in the field depending on when it failed. The failure mode at the dispensing station would be "excessive adhesive applied" and the cause would be related to excessive air pressure or too large of dispensing orifice or insufficient viscosity of adhesive and so on depending on what variables determine the proper dispensed amount.

At the calibration station, the failure mode would be depending on where the crack occurred something like 'saturated output' or 'no output' or 'out of spec part' and the cause would be 'cracked hybrid' while the effect would be an increase in scrape. Same at End of Line testing if it made it that far.

When it comes to rating these conditions, that would depend on the industry, individual customer or even your own company. We generally followed by the guidelines dictated by the automotive industry which provided the rating system. Even within using their system it is alot of guess work because depending on how they use your product will depend on what you rate the severity.

If a pressure sensor goes bad does the customer have their vehicle go into a 'safe' mode where you can keep on driving but will need to get the issue fixed or do they have the vehicle come to a dead stop? If our case severity was dependant upon how they programed their ECU module.

In reality as long as you consistantly rate everything the same way then the issues that are most important will have the highest RPNs whether they are at 125 or 75.

Vadim,

Speaking from experience, FMEA analysis on a design is a valuable tool. For us, it opened several AH HAs for improving our design.

It is unfortunate that your boss does not want to invest in training for such a valuable process. Perhaps he will entertain hiring a consultant to come on board to get you through at least one iteration of your FMEA.

Function: Perform a FMEA

Failure Modes:

1. Product unreliable
2. Poor customer satisfaction/Loss of sales

Causes of Failure Modes:

Causes of 1:

1. Failure mode oversight
2. missed reliability factors in design

Causes of 2:

1. Loss of Customer confidence