I find YouTube to be a good source of information on many subjects. Want to know how to fix a widget? There are probably a few videos on that exact subject.  Technical subjects are also there.  I was searching on “FMEA” and found an interesting looking video on the “Top Ten FMEA Mistakes.” As I was working on a team to improve the design FMEA process for functional safety, I clicked on it. On the title slide, I noted that this YouTube video was made from a portion of a presentation made in 2006 by Richard A. Harpster of Harpo Systems, Inc.  I first thought it might not be relevant as it was old but proceeded as it had 66 thousand views!

Among the top ten mistakes were many of the same things I have heard from exida certification customers when I recommended DFMEA as the architecture design verification method in their development process. The top four items matched my list!

1. FMEA responsibility from the Quality Department not the designers. DFMEA is a design verification tool not a document creation exercise to comply with a contract. The design team must be responsible. 
2. Too many, wrong people at the analysis. A complete cross functional team is not needed. For DFMEA, a much smaller group focused on the design and potential design improvements is much better (and costs far fewer engineering hours).
3. Wrong time. DFMEA provides the benefit when done as THE design verification step in the development process, done as part of the architecture design.  Not done later to fulfill a contract date when design problems might be rationalized away as changes late in a project are very costly.
4. FMEA not integrated with the development process. DFMEA is clearly most effective when used as part of the design process - including iterative processes. Testing objectives including integration test planning and FMEDA info documentation can and should be part of the DFMEA.

All these mistakes lead to higher cost and lower value. I have heard from some engineering mangers and designers that DFMEA is not the worth the cost. I can understand that point of view as some DFMEA methods burn up a lot of engineering hours. But the DFMEA method can also save a lot of engineering hours in rework later in a project. That has been proven over several decades.  

A DFMEA method that is integrated into the development process can reduce engineering time and project schedule by doing what FMEA does best, reduce rework. With enhancements for functional safety and elimination of duplicate work, the process can do even better.  Additional information that should be addressed in a DFMEA includes:

1. Description of automatic diagnostic functions,
2. Action taken when an automatic diagnostic detects a deviation from proper operation,
3. Estimate of automatic diagnostic effectiveness for each deviation (functional failure mode),
4. Description of any Latent Fault testing (start-up tests, operator periodic tests, etc.),
5. Estimate of Latent Fault effectiveness for each deviation,
6. Integration test objective to verify any proposed design mitigation, and 
7. Design justification documentation.

These things are needed for a functional safety development process. These items are also needed to do an FMEDA.

I do not see any of this additional information needed for functional safety in any DFMEA standard/handbook/textbook. Of course, that is understandable since the DFMEA methods were created long before complex microcomputers and functional safety principles existed.

A better DFMEA method is needed, especially for functional safety. We call this method Design Deviation and Mitigation Analysis (DDMA). The value comes when DDMA is used as the architecture design verification step in the development process. RPN numbers could be used but not needed – that saves time. More time is saved in the FMEDA  as needed information is gathered when the design team is in the context of the architecture design.

Next generation DDMA tools add “Knowledge Bases” of well-known failure modes for each unit type called Deviations and proven Mitigations.  More time saved.  With all these savings, perhaps engineering managers will be much more receptive to the recommendation of DFMEA / DDMA.