FMEA Severity Rating for aggravating failures

[cfilion]

Hi,

We manufacture complex products.
If a process fails it may or may not lead to a problem down the road.

Ex: a poor weld may create or not a sharp edge that could wear a part close by and initiate eventually a short-circuit.

I know the frequency rating of poor welds.
I know the worst outcome possible (loss of the unit) if the conditions are all wrong: poor weld with a sharp edge, high vibrations, weak cable protection ... very low probability.
I know the Detection level for this defect.

From the above how do I select the severity ratings for poor welds?
I see the poor welds as a contributing factor or an aggravating defect but not the direct cause of failure! (In aerospace I would say not a direct path to strain)

Any idea?

Thank you

 

[John C. Abnet]

Severity has to do with the consequence that will be experienced IF the failure mode is realized.

Severity does not "care" about the potential that it will occur (frequency or likelihood that it will be detected).

Severity simply considers IF the failure mode does indeed occur, how severe is the consequence (based on the scale of severity you are using).

Hope this helps.
Be well.

 

[Miner]

Also, you should only consider direct effects of the failure mode on the customer, end user or downstream operations. Many engineers get trapped by considering potential "butterfly effects" of a failure mode. For example, a direct effect for an end user might be an increase in stopping distance for their vehicle. This is a legitimate effect. A butterfly effect would be that the increase in stopping distance resulted in a 40-vehicle pile-up. This is not a legitimate effect.

 

[Bev D]

I would add to the great responses above that we can’t disconnect causes from the failure mode(s). All failures have a cause - they do not just randomly occur for no physical reason.

This is exactly why the causes of failures are part of the FMEA. To reduce or eliminate the failure we must reduce, eliminate or compensate for the cause. The only exception to this is where we error proof the failure itself in some way. (Example: I worked on a team charged with solving a failure to eject a one time use object from a system. Fixing the cause was going to be incredibly expensive, the team came up with a software driven fix to add a mechanical repositioning of the object went the system detected a jam. A retry to eject was then performed and it dropped the failure from 3% to less than 0.1%)

 

[cfilion]

I love the butterfly effect reasoning.

When can we exclude such an effect?
In my case, I will apply as if the poor weld will affect the functionality of the product, to be on the safe side.

I would like to explore another way of looking at this.
1) The impacts of generating this defect as a function of various level of customers
ex:
The severity for the next operation in production is = Will need to rework (send back)
The severity for Customers is = May lose functionality of insulation (the part being in their hands)

2) Then for Insulation
The severity of losing insulation: May lose functionality of the product.

All with their own O (occurrence) and D (detection potential)

Although I am not ready yet to defend or expand the various customer-level approach, I find that it has the potential to indicate other ways of reducing the impact or the likelihood.
Similar to the above example of using a software fix for a complex process, by using a customer level approach we resolve at the level that makes the most sense.

 

[Miner]

Let's try another example with which I am more familiar. Take a motor/gearbox assembly used in pairs on electric wheelchairs.

One potential failure mode is that the motor to gearbox interface becomes disconnected over time. The immediate effect would be that the motor could not drive the gearbox. Since this does not happen until the electric wheelchair was in the field, an effect on the next operation or on the customer does not exist. The effect on the end user would be that the wheelchair could no longer move forward, but only spin in a circle. This is where you stop and assess the severity. Going further would be to generate butterfly effects such as the end user becomes stranded and dies from exposure, or the chair spins out of control going down a ramp throwing the occupant from the chair.

 

[Bev D]

I have always considered teh “Customer“ in an FMEA to be the end user. Or I have used bot hmy immediate Customer and the end user. The end user who is most important.

I also use a local and system effect to avoid the diversionary term Customer…and in my severity descriptions I include operations, legal & regulatory and totl Customer. In my last organization that included the vet as well as the patient…

Miner is right about the butterfly effect. I’ve always called them the zombie vampire meteor effect…usually used to mock the FMEA process. Some butter fly effects are viable as they can be terms of lawsuits. However if we are diligent regarding the severity of the physical faiure and don’t minimize occurrence by guessing low we will reduce, prevent or control teh cause so as to actually minimize failure occurrence.

Remeber the purpose of the FMEA is to improve product quality…unless you are doing it to check box.