How to Maintain the FMEA

      FMEAs are initially established in the early stage of product/process design and development. They're living documents and should be regularly reviewed and updated. Normally, FMEA review and updating should be triggered by the following events:

1. There're changes in the product/process requirements
      If there are changes in the product or process requirements, the FMEA team should review the changes and decide whether these changes can result in any new failure mode. If yes, document these new failure modes and complete the rest of FMEA accordingly.

2. A new failure mode is detected
      When creating FMEA, the FMEA team should list up all the failure modes that can occur to the product/process. Due to the limit of the team's knowledge, however, this list in most cases is not a complete list. Later on, when a new failure mode is detected, e.g. during design verification and validation, during mass production or during customer application, it should be included in the FMEA.  

3. A new cause is identified for a failure mode
      When creating FMEA, the FMEA team should also list up all the possible causes to each failure mode identified. Again, due to the limit of the team's knowledge, this list is often not a complete list. If a new root cause is identified in the failure analysis later on, it should be included in the FMEA.  

4. There's updated information to evaluate the ranking of S, O and/or D
      When creating FMEA, the FMEA team can sometimes only do the risk assessment (i.e. ranking of S, O and D) based on the best guess, as objective data or evidences are not available yet. With further deployment of the design and development activities, mass production and customer feedback, more and more objective data and evidences can be collected for more accurate ranking of S, O and D. When such information is available, the FMEA should be reviewed to re-evaluate and, if needed, to update the SOD ranking. For example, it is a common practice in factories that failure rate and cause of occurrence of each type of defects is consolidated on a monthly basis. These data should be used for the ranking of O, e.g. if a certain cause can result into a monthly failure rate of 0.2% for a certain type of defect, it indicates O=6 for this specific cause (according to the scoring criteria of AIAG's FMEA manual). If O is scored otherwise in the current revision of FMEA, it should be updated to reflect the actual situation. 
      If recommended actions are proposed, the FMEA should be updated later on to document the results of the actions. This should also be done after the actions are completed and objective data are available to evaluate the results.

5. The preventive and/or detective actions are found not effective
      As mentioned above, the risk assessment in the current revision of FMEA may not be accurate due to the limit of objective data. When these data are available, one may find out that SOD are severely under-ranked and the preventive and detective actions implemented are actually not effective. In such cases, more effective actions are should be recommended and the FMEA should be updated accordingly.

6. Proactive improvement is initiated
      One of the key principles in quality management is continual improvement. Following this principle, organizations should proactively initiate some improvement activities on a regular basis. FMEA is a good source to determine what kind of improvement activities can be initiated. It is a common practice that the FMEAs are regularly reviewed (e.g. annually), and top risks in the current revision of FMEAs are picked and actions are taken to further reduce the risks  (Please be noted that the current controls for the top risks are effective and the actions are not mandatory. Also please refer to this article on how to pick the top risks). These actions should be updated into the FMEA as the recommended actions, and the action results should be documented in the FMEA as well. 
.

How to Determine Whether Recommended Actions are Needed after the Initial Risk Assessment with FMEA

    AIAG’s FMEA manual (P.57 and P.103 in 4th edition) requires that once the initial risk assessment is done in FMEA, it should be decided whether further efforts are needed to reduce the risk. In the 3rd revision of the manual,  a threshold of RPN=100 was set as the criteria, above which actions are needed, and vice versa. In the 4th edition, however, it is recommended NOT to use such a threshold to determine the need for actions, as it is against the concept of continual improvement (If a threshold is defined, no further actions will be taken if the RPNs in an FMEA are all below it). The 4th revision of FMEA manual now requires that the FMEA team should do review on their FMEAs and continually take new actions for improvement.

    Now let’s see how FMEA review should be done and in which cases new actions should be proposed.

    First, review should be done shortly after the initial establishment of the FMEA to identify all failure modes with S=9 or 10, as required in the 4th edition of the FMEA manual (P.103).

For these failure modes, actions should be proposed to reduce the risk. For example, the design team may change the design to reduce the score of severity until S≤8. If change of the design is not possible, the design team should then propose actions to reduce the score of occurrence and detection. How much should they be reduced? It depends on the customer requirements. For example, I have seen customer which requires that occurrence must be reduced until O=1 for any failure modes with S=9 or 10. In such cases, the FMEA team needs to think of some error proofing methods to eliminate the occurrence of the failure modes. If no customer requirements exist, the organization can determine internally their requirement for occurrence and detection when S=9 or 10. A reasonable requirement is to reduce the possibility of occurrence until O≤3, as most automotive customers require Cpk≧1.33 for special characteristics (Failure modes with S=9 or 10 should be regarded as special characteristics due to safety concern. Cpk≧1.33 is equivalent of a PPM≤63, O=3 is equivalent to PPM≤10 and O=4 is equivalent to PPM≤100. To meet the customer requirement of Cpk≧1.33, O should at least be 3). The above review and actions should be ideally taken before the release of product or process.

    Once the actions have been proposed and taken for all failure modes with S=9 and 10, and have effectively reduced O and D to the desired level, the FMEA team should now do review on failure modes with S≤8 with a regular interval, e.g. once a year, after the release of the product and process (such review is not mandatory before product or process release). Since there are so many failure modes in the FMEA, it’s not possible to propose and take actions for all of them, the organization should establish criteria to pick the failure modes for which actions will be proposed and taken. The 4th edition of FMEA manual requires that those failure modes with the highest O or D should be picked (P.103):

Other than that, no further requirement is specified in the manual. The organization has its own freedom to set up more specified criteria (but of course customer requirements should be followed if there’s any). To limit the work load within a reasonable scale, the organization may use the criteria below to pick out the failure modes with S≤8 for further improvement:

1. The team should first pick out the failure modes with highest O or D (no priority between O and D). 
2. If there exist more than 3 failure modes which meet the above criteria, the team should pick out, among them, the ones with highest RPN.
3. If there still exist more than 3 failure modes which meet the above criteria, the team should pick out, among them, the ones with highest SxO or SxD (no priority between SxO and SxD).  
4. If there still exist more than 3 failure modes which meet the above criteria, the team should pick out, among them, the first three as appeared in the control plan.

    Let’s take a look at one example with 8 failure modes as shown below. In this example, the FMEA team first picks out No. 3, 4, 5, 6 and 7, as they have the highest O or D=8 (there’s no priority between O and D). Among these five failure modes, the team then looks RPN, and picks out those with highest RPN=240, which are No. 3, 4, 6 and 7. Among these four, the team then looks at the ones with highest SxD or SxO (there’s no priority between SxO and SxD), and picks out No. 3, 4 and 6. At this point, there’re only three failure modes left, then the team should propose and take actions for them for improvement.

    As summary, when there’s no customer requirement exists, the organization may establish such criteria to review FMEA and take actions for continual improvement:

1. Review the FMEA shortly after its establishment and before the product and process release and take actions for all failure modes with S=9 and 10, to reduce the occurrence until O≤3;
2. Review the FMEA once a year after product and process release, and follow the criteria below to pick out no more than 3 failure modes and take actions for improvement:

• Pick out the failure modes with highest O or D (no priority between O and D). 
• If there exist more than 3 failure modes which meet the above criteria, among them pick out the ones with highest RPN.
• If there still exist more than 3 failure modes which meet the above criteria, among them pick out the ones with highest SxO or SxD (no priority between SxO and SxD).  
• If there still exist more than 3 failure modes which meet the above criteria, among them pick out the first three as appeared in the control plan.