The FMEA Mini-Series: Using an FMEA vs. SxO for Prioritizing

I'd like some comments on occurance. The tables in the manual I've always had interepretted in PPM or in terms of the number of parts. When doing a process FMEA and looking at a process failure, is it also how often the event may occur? For example if there could be a set-up problem that you think occurs 1/20 set-ups, but only 1 piece out of the schedule run of 100,000 gets produced because 1st piece inspection catches it, is occurrance 1/20 or 1/100,000?

I have an opinion, but am interested in hearing some others first.

Well I'll step out on the limb here and say, IMO; that "occurrence" means "What is the probability that the potential effect will occur, not how many times it will occur" This can only be established by determining the confidence limits of your process, i.e., .90, .95, .999999.

-> This can only be established by determining the
-> confidence limits of your process, i.e., .90, .95,
-> .999999.

How about for a design FMEA?

I think in Laura's example, the set-up would be a "process function" and not related to occurance. Occurance is tied to the projected rate that the potential cause will occur.

Hopefully there is data (PPM, Scrap, Rework etc...) from past or current projects that can be used in the determiniation of your ranking of the occurance.

When thinking occurance don't consider effects or controls, think of causes.

ASD...

In Laura's scenerio, I take the stance that the occurrence stays at 1 in 20.

However, the first time inspection is the DETECTION, isn't it? Therefore, if the first time inspection is robust, the overall RPN will be driven down. So, in and of itself, the occurrence of 1 in 20 is not necessarily a bad thing. But it is a huge red flag to make sure that BOTH your detection is robust, AND make sure that your severity is reasonable. If you have a high severity and a 1 in 20 occurrence, I would think that the process may need additional tweaking.

HM

I tend to agree with Al. My first question is "What is a failure?" This may determine how you treat occurrence due to the specific cause. If you are addressing a specific product failure mode, then I would take the position that the occurrence is 1/100,000, not 1/20. You have 1/20 setups fail that result in 1/100,000 failed products. You can also think of it in terms of Cpk. Would you say your production process is incapable because your setup fails 1/20 times and generates 1/100,000 failures? I wouldn't. (I would still go fix my setup process anyway because a setup part still has $$$ associated with it.) I would evaluate my baseline process capability outside of any setup to determine occurrence.

JRS

Energy –

I think we are arguing semantics here. The FMEA manual states “The possible failure rates are based on the number of failures which are anticipated during the process execution”. In my own opinion the setup process is part of the overall global “process execution”.

Look at the broad picture as Al suggests. Base the occurrence value off historical performance data such as PPM, scrap or fallout from past or current projects.

I have watched many FMEA meetings get hung up on these types of questions. If no historical data is present, play is safe and score to the high side. Keep the FMEA alive, build some history and update as more data and history becomes available.

I am having a bit of trouble understanding this, I guess.

There are some assumptions being made by everyone regarding Laura's example. And, I think we maybe aren't making the same assumptions.

So, help me out Al and J.R. In your interpretation of the scenerio, what would be the process weakness, and what is the process control?

I guess I am looking at it as if the set-up is the weakness, and the inspection is the control. And I think of inspection as a type 2 control. The inspection doesn't really change the occurrence of a wrong set-up. It does, however, detect it.

It almost seems like you are considering the inpection as a type 1 control. Maybe that is acceptable.

Just different ways of looking at it, I suppose. I trust that you will let me know if I am off base.

HM

I guess I would look at the situation this way:

1: Process Set-Up Is A Process Funtion.

2: Oversized Diameter would be a Potential Failure Mode

3: Unable to process through next operation would be the Potential Effect Of The Failure.

4: Lack of set-up instructions would be the Potential Cause Of The Failure.

5: For Current Process Controls, setup instructions would be a Prevention Control, and 1st artical inspection would be a Detection Control.

The Occurance value would be derived based on #4 from above, the "Potential" occurance of the cause.

This is an over-simplified example.

Severity --- Potential Effect
Occurance --- Potantial Cause
Detection --- Current Controls (Prevention & Detection)

In Laura's example I don't see a reference to the cause of the failed set-up. There might be multiple causes which would call for multiple occurance ratings.

MHO

ASD...

So would it depend, then, on whether it is a set-up dependent(machining) or operator dependent(assembly) operation? APQP has alot of examples of different "types" of processes, but the FMEA does not.

My thoughts aligned with ASD - however the company did not want to be as specific as "would not assemble" as the effect. They preferred "parts out of spec" - because they are a machine shop and didn't necessarily know the detail of the effect. So each dimension wasn't specifically identified as out of spec - the entire operation was grouped as "parts out of spec." The customer rep was actually present, but also didn't know all of the effects, and the group really wanted to focus on eliminating the cause.

MHO? - the occurrance criteria being used should be established when you start out (we didn't do that). That the PFMEA manual is intended to be in terms of parts (rejects, PPM) and that it needs to be "changed" by company procedures if the processes are set-up dependent. I started out thinking 1/100,000 but changed to 1/20 - with the number off the table being the same - just modifying the criteria.