Residual Risk

In a FMEA you often encounter multiple actions that are planned and executed to address a single cause of failure. Such actions may be planned in a single session, or may span over a period of time. The FMEA format recommends that as a result of each action taken a closing Severity Occurrence and Detection ranking is logged. There is also a column to log the closing RPN.

I take these closing numbers as the status of the risk after the action has been implemented. I do not know how the Severity ranking could change as a result of an action, so I prefer to leave it as identified with the most serious effect of the Failure Mode.

I believe the closing Occurrence and Detection rankings tell me something about the residual risk that remains after taking the action. With multiple actions implemented there will be several pairs of closing Occurrence and Detection rankings each indicating the residual risk that remains after the action is taken.

To understand the residual risk for the Cause to lead to the Failure Mode after all actions, I look at the lowest closing Detection and Occurrence rankings across all the actions completed to address the cause. There isn’t a lot of technical literature that I have come across that talks of the residual risk assumption as a result of multiple actions put together.

Comments and discussions from practitioners would help. Thank you.

 

Yes, all that's pretty standard. I can envision cases where you could lower the severity (consider throttling a car to only reach max speed of 30 - not very practical but would likely reduce severity of crashes!). It's pretty uncommon in my experience and I typically don't consider changing it post-controls.

For any individual risk, the residual risk is just considered for the particular mitigations to be implemented. The standard (14971:2012) states that a risk-benefit analysis needs to be conducted on each one to determine if the benefits outweigh the risks. I typically look at industry standards initially and then, after deployment, start looking at feedback, complaints, adverse events, etc. to support whether the benefits, in fact, outweigh the risks.

The standard also requires that the risk-benefit analysis conducted at the product level; i.e., do the benefits of using the device outweigh the "sum" of the residual risk. Before distribution, this is a pretty weak guess. After distribution you at least have some data. One of the better approaches to this that I've seen was using a heat map. If all the post-control RPNs are clustered near your "red" zone, there may be some cause for concern. I especially liked the illustration given: an airplane with a bunch of "bandaids" covering small cracks. Individually, maybe none represent a big concern; but the number of small issues when taken together would clearly be a concern.

One could, in theory I suppose, do a heat map for particular failure modes. I don't know, though, if looking at just the lowest RPN gives you any indication of the effectiveness, though. Maybe I'm not understanding the approach you're considering?