FMEA Detectability

Six Sigma – iSixSigma Forums Old Forums General FMEA Detectability

Viewing 8 posts - 1 through 8 (of 8 total)
  • Author
  • #48659


    I’m a design engieer for an implantable medical devices (ie high risk).  We use a Design FMEA on every new product or product change.  We have team members from engineering, marketing, clinical/nursing and regulatory – and the development teams get confused about Detectability almost every time.  Many of them seem to think “detection” is an absolute (like Severity) and does not change based on the mitigation.  I have been working to better define Detectability in the FMEA.  I’d like some feedback from this group.
    Design FMEA – how detectable (think preventable) is the Cause of this failure (this line on the FMEA table).  For example, if a potential failure cause is a poor weld, and welding is a new process for us, what will we do (mitigate) to prevent this before the product is released to production.  After mitigations are implemented (and we know they work) the final detectability rating takes into account the mitigation.  That is, “given the mitigation, how detectable is the cause of the failure”.  If the mitigation to the weld example was to perform a weld section-inspection on samples from each production lot, the post-mitigation RPN rating should be reduced by both a reduction in Occurrence and Detectability.



    I feel your pain with this, as many others at the forum, it seems to be almost weekly discussion. My only advice for this situation is to simply have a short training session before actually beginning the FMEA process. Since you obviously have access to previous successes, use them as examples.
    A source you may want to visit is the AIGA standards-definitions are so so, but may be of help.



    This is the guideline document for the auto industry right?  I think I have read several articles/examples regarding this.  As I remember, it uses a more of an action item approach to mitigation – listing recommended actions to reduct the RPN.  I like this but can’t use it for two reasons.  In our regulated industry we are all about “formally closing the loop”, that is being able to show exactly what was done with proof that it was effective (thus the need for post mitigation detection ratings).  The second reason is institutional inertia – the guys that wrote the current FMEA work instruction are still running the company.  Adding a better definition of Detectability will be a hard enough first step. 
    Sorry if I’m wrong about the AIGI, but I will review the document.  Thanks. 



    In your example, increasing appraisal activity could potentially impact your detectability score (although not my first option) but would not have any effect on your level of occurance.  Obviously, you are looking at inspecting out nonconformance with this approach, and it should not impact any other category other than detectability. 
    If these testing and inspection procedures lead to preventative actions, then at that point you could adjust your occurance rating. 
    Good luck.



    That was not the best example since it overlapped design a production.
    Let me try a better example. Early in the design phase the team realizes that the mechanical fits between parts in a particular assembly is critical to the function of the assembly.  Since there is no in-place design control to prevent the root cause of poor design tolerance allocation  (except for how good the design engineer is) O is rated high and D (detectability, ie how detectable/preventable is this situation in the design) is also rated high.  Next a mitigation is planned – a formal report of the tolerance stackup on the parts.  Once the Stackup Report is complete, which shows if the parts are built to print the assembly-fits should function, this fmea line item can be reviewed again.  Given the Report, the consensus is that now the occurrence the “poor design of the assembly fits” is very unlikely (O is rated low) and now, given the report,  the detectability/preventability of poor desing is also low.  Before the report the detectability/preventability was high, with the report the detecability/preventability becomes low.  All of this happens in the design phase, before the first units are even put together for verification or production.
    We know this type of problem would be found during integration testing of the new product – perhaps a year after the design started, but that is too late.  In our Design FMEAs we are trying to identify elements in the design that are likely to cause problems during verification testing, costing more money and time to fix, as well as trying to reduce risks to the user. 



    Roger,With all due respect, I don’t think you understand occurance or
    detection. Some silly report doesn’t reduce occurance, proven process
    capability can. The silly report for sure does not reduce detection,
    detection is about how certain we are to know a problem has occured,
    it has nothing to do with some tolerance stack up. Detection – think
    mistake proofing, process control, tests and inspections in that order.



    That makes sense with a finished product, and for manufactuing processes, and those are the definitions we use for an Application FMEA (currently marketed device being assessed for a new use/new medical application).  However, the FDA requires – by law – a Risk Assessment prior to market release.
    See the link:
    This is just one of the first links on a search result on the terms “fmea fda”.  As you will see there is no shortage of consultants working in this narrow nitch.
    For a medical device, say a pacemaker, if the problem occurs, ever, the patient may die.  What the FDA wants to see is “detection” of potential design problems during the design phase.  How can you mistake proof the design process itself?  As the FDA says with respect to a product design – test is not a mitigation for design weakness.
    You can see why we struggle with this.



    Hello Stan,
    Just one comment.
    I look at the detection phase as the ability to detect a mistake that has already taken place (stop it from being passed on to the Customer).
    I think the detection is independent from the occurrence of a defect.
    As an example, a process could produce 50% defects but 100% of the parts are thoroughly inspected and as a result, no defects are passed on to my customers.
    If you have a ‘mistake proofing’ system in place, you prevent ‘wrong parts’ to be produced in the first place and this implies that occurrence will fall to ‘very improbable’ and may be, inspection is no longer required.
    Do you agree with that ?
    PS: Anyone’s comment is welcome

Viewing 8 posts - 1 through 8 (of 8 total)

The forum ‘General’ is closed to new topics and replies.