FMEA - Current control and occurrence rating

[WS Kang]

Hello

I have some questions about fmea.

1. All current control (prevention and detection control) should be executed?
I think all current contol should be executed becase i should verify effectiveness of current design.
I see in the recommened action step, all action is contolled and monitored visually in the fmea sheet But regarding current control action (date, responsible person, result) there is no comumn in the fmea.

2 As i know Occurence level is affected by prevention contol and detection contorol is affected by detection control. Initially i might rate occurrence and detection level based on planned or progressing current control
At this time , my rating is just estimation because there is no history or evidence.
Suppose after i perform test as a detection control for actual design from prevention control, and no issue is found. In this case can i correct initial occurrence rating (column next to current control) not column after recommended action because i feel more reliable for current design? If so, occurence score will be reduced without any design change. Is my thinking correct?

WS kang

 

[Bev D]

There are several issues with FMEA that result for misunderstanding it’s intent and how some (many) organizations implement it as a check the box form.
Without discussing the meaninglessness of the RPN value, let’s address the detection and occurrence ratings.

Detection can really only be ‘rated’ with data. That means you must perform the necessary MSA tests for everything from an error proofing control to a human visual inspection. Without this you are only guessing.

Occurence can really only be rated after you have tested the product or process at it’s worst case input tolerance limits and worst case use conditions. You could use random samples that span the range of inputs but that would mean very large sample sizes. You must have data to rate the occurrence or else you are only guessing.

The point of FMEA is that you start at the beginning of the development process. The form is populated only with functions. You should have a lot of questions. Severity and potential causes can be determined by your knowledge of science and physics. Then you begin characterizing and developing your design by running screening experiments to determine the critical input characteristics that control your critical outputs. You then proceed to running experiments that determine the ‘design space’ or what limits you need on the inputs to guarantee your outputs meet specifications and requirements. At the same time you develop and test control methods that maintain those input tolerances and test the outputs to ensure they meet specifications. At the end you will have a completed FMEA.

This is an iterative process. The FMEA is guide to keep you on track. Think about it this way: Severity is an input. Occurrence is an output. Detection is a Control.

Those who fill out an FMEA form at the end of the development process are wasting their time. It’s too late to be anything but a frustrating exercise in filling out the form and scrambling to bandaid the design misses.

 

[WS Kang]

Thanks.
I agree with your opinion. But still have some confusion. At the beginning occurrence level might be relatively high due to not enough field or reliability data. And later reliability test(current detection control) is competed which means detection control is completed and now i have data which prove my design. So i can reduce occurrence score, right? In this case i can reduce occurrence level without design change. If this this situration happen, should i correct occurrence score which i initially wrote not in the final score? I think this is not design change so basically occurrence is not changed. Or should i put it in the cell in the recommended action? I think there is no reason for this because no current control method is changed and just current control is impemented. I would like to make it clear.

 

[Bev D]

IF your validation/reliability testing indicates that the occurence level is smaller than what you originally guessed THEN you should reduce the rating to the appropriate level. There is no requirement that says that occurence can only be reduced by a design change. Of course the best approach is to not enter any occurence until you actually know what the occurence is through testing.