
AI Future of FMEA
Abstract
Carl and Fred discuss the use of AI with FMEA, what works and what does not work.
Key Points
Join Carl and Fred as they discuss the broad subject of FMEA and AI. We look at early lessons learned, discuss opportunities, and focus on the most important concerns.
Topics include:
- How FMEA can interact with Model-Based Engineering?
- FMEA pre-population, which fields can be pre-populated and which should not?
- What is role of human team in FMEA?
- What are primary ways to reduce FMEA in-meeting time?
- How do team-based FMEAs, done very well, compare to AI automated FMEAs?
- How can AI augment well-done FMEAs?
- What specific support can AI provide to FMEA?
- AI can review past problems and provide excellent input to new FMEAs
- Look for areas of concern by proper FMEA team
- One concern about pre-population: people don’t see what isn’t there
- Consider AI bot as FMEA team member, with specifically defined role
- AI cannot replace human creativity and ingenuity
- Part of the value of FMEA is the deep discussion between subject matter experts
- AI will not solve potential blind spots
- AI can help with augmentation for FMEA, but not taking over the FMEA process.

- Social:
- Link:
- Embed:
Show Notes
Hello,
We are currently working to establish some formal reliability processes at my company because much of what happens is up to which reliability engineer being asked at the time.
One of the suggestions put forth is to stop using MTBF requirements (hooray!) but the replacement would be basing requirements off of RPN and Item Criticality from a FMECA. This does not make much sense to me. What are your thoughts on this approach? I advocated for using actual reliability as a requirement but gained little traction. The company is a defense systems integrator/designer.
Hello John.
Glad to hear MTBF is dropped as a metric for reliabilty requirements.
Unfortunately, RPN is not a viable metric to derive a reliability number, the reason is the constituents of RPN are subjective assessments, and reliability requirements are objective.
Regarding use of item criticality from FMECA, this depends on the level of objective accuracy in calculating item criticality. If the failure rate estimate that goes into the item criticality calculation is valid and not subjective, this can be considered. However, when doing FMECA using quantitative criticality assessment on new designs or new applications, often the failure rate estimates are quite subjective, even if informed by field data for similar products. This is due to changes in the new product being analyzed.
There are better ways to develop good reliability requirements. Building them from FMEA or FMECA has pitfalls.
You can search on Accendo Reliability for information on creating viable reliability requirements. I’ll also be podcasting on this in the near future.
Carl
Hi Carl,
I appreciate the reply. I looked at other resources and articles on Accendo Reliability for reliability requirements and found them to be helpful. These other articles and resources are what I based my recommendations off of, but I am afraid we are going to continue to go down the wrong path chasing the RPN. I wanted a sanity check more than anything. I look forwards to your podcast on this subject! Thanks!