FMEA is a method of assessing what could potentially go wrong with a machine/system and devising actions that will eliminate the possibility of a failure happening or at least reduce the risk to an acceptable level.
I am no expert on this topic, but have been involved in FMEAs at my work. We have assessed designs for new equipment that have the potential to kill people if things go badly wrong. Often the engineers end up adding layers of protection, so that if a critical component fails, there is a second or even third system that will kick in to prevent catastrophe. Hopefully your robot will not kill someone if it fails, but the methodology is great for improving reliability. At Worlds, I have witnessed a number of robots have failure of critical components that have resulted in “game over” (including our own robot :rolleyes:). A documented FMEA would be a useful addition to a design notebook.
Basically what you do is:
- Consider each subsystem of your robot, for example start with the Drive Train.
- Think of all the possible failure modes - stalled motors, stuck on bump, tipping over, broken chain, axle comes out of motor, etc.
- Document what the effects could be - temporarily immobilised, reduced drive speed, permanently immobilised, etc
- Give each failure mode a score based on probability of it happening and the severity of the effect if it does happen. High score means it is likely to happen, or it is “game over” if it does happen, or both!
- If you gave it a high score, do something to reduce the risk - gear down drive speed, raise chassis height, add anti-tipping mechanism, etc
Some things you can do nothing about and are just bad luck, but most things you can do something about to help prevent them from happening. If you are interested, you can find more info here.
http://en.wikipedia.org/wiki/Failure_mode_and_effects_analysis
Cheers, Paul
Paul,
Just curious, which type of FMEA approach did you use with Robotics (automotive style FMEA using Risk Priority Numbers (RPNs) or the defense/aerospace approach (Mil-Std-1629A) that uses Criticality Numbers?
Thanks!
Michael
Hi Michael
Wow, this was an old post. 
To be honest, my powers of mentoring have never reached a level where I have been successful at influencing anyone to carry this out in a formal, written fashion. However, this methodology has been used in discussion form regarding robot reliability and reducing/eliminating potential robot failures with the teams I have been involved with. It is a good way to prevent the predicted outcomes of Murphy’s Law from occurring.
A difficulty for newer teams is that initially, they are unfamiliar with the various VEX components, so they don’t know what they don’t know regarding what might fail and what is likely to fail. This makes it harder to come up with the potential modes of failure.
I had a very quick read up on the two approaches of FMEA that you mentioned. My honest opinion is that both may be more complex than is required for VEX students. I would simply use Probability and Severity to give a ranking score for each failure mode. Then come up with the mitigating actions, one of which might be as simple as doing a weekly maintenance check that all the screws are still tight.
Torqueative pretty much nailed it in the head. Having worked in an aerospace / defense company, I can second the comment that the critical number approach is quite extensive and thus complicated. It is quite useful after the critical numbers are assessed, especially in gauging relative probability of failure modes, however that’s not a simple task.
I would recommend the probability & severity approach as mentioned by Torqueative. The product of the two would be the severity and that surprisingly will tell a lot about how the robot (or even team) will perform. During my undergrad senior design project, this approach was taken in the name of risk management, and just being able to anticipate what can go wrong and plan for it can go a long way.