Help: detection rank in FMEA

Six Sigma – iSixSigma Forums General Forums Tools & Templates Help: detection rank in FMEA

This topic contains 8 replies, has 6 voices, and was last updated by  Paulonis 8 years, 4 months ago.

Viewing 9 posts - 1 through 9 (of 9 total)
  • Author
  • #53801


    Hi, Folks,

    when we rank the detection score in FMEA, do we rank against the detection of the effect, or rank against the root cause that causes the effect?

    For examle:
    Process: toast bread
    Effect: Bread burn out
    Failure Mode: long baking time causes bread burn out
    Root cause: wrong time setting

    If we rank the detection against the effect regardless the root cause, then Detection could be 1; but if we rank against the root causes that cause the effect, then the detection would be higher, let’s say 8.

    Which one makes sense?

    Thanks in advance for help!



    Think about it. Each potential cause has its own number for likelihood so the combination of cause and effect is already part of the RPN calculation.



    Detection is the licklyhood that you will spot the problem before the client with 1 being you would not and 10 you would alwys do this. It has no link to the effect or the root cause.

    So in your example.
    If you are making it for your self and the only time you know it’s burnt is when it’s burnt then that would be a 1.
    If you were making it for someone else and you can see it’s burnt when you take it out of the toaster before you give it to the person than thats probably a 10.



    As mentioned, the detection aspect is your ability to detect the defect. It is based on your sampling and testing methods.

    One slight issue with the previous poster. The detection score is “flipped” versus the frequency and severity scales. Your detection score is 1 if you will always detect, and the maximum value (3, 5, 10, or whatever range you’re using), if you will never detect the defect. This way, as your ability to detect improves, the issues importance decreases.

    As always with FMEA, make sure you’ve got the scoring grids well understood before starting the process.



    **Slaps head**

    Sorry about that should read what I write before I post….. They should be the other way round.



    I score and teach detection as the ability of the control(s) to detect *and prevent* the impact of the failure mode from reaching the customer of the process output. Detection implies that action could be taken such that the effect of failure does not reach the “customer”.

    The detection score includes consideration of where the control(s) impact the causal chain. Typically, the earlier in the chain, the lower the score. This is not an absolute rule, but controls that focus more on early prevention tend to be more effective than those based on last-minute inspection. In most cases, I would give a control based on SPC of process inputs a lower detection score (better detection) than a control based on SPC of process outputs, which would in turn get a lower score than inspection of the final product at shipping. Again, this is not a hard and fast rule, but is one that I give significant consideration to.

    There may be some disagreement of how to score mistake-proofing controls that are purely prevention. Some would consider those to affect occurence rather than detection. In practice, there isn’t a lot of difference in the end result with either approach, as RPNs that include a 1 in occurence or a 1 in detection are unlikely to rise to the action level for most FMEAs.



    I agree with Paulonis, the best detection is prevention. Detection before an operationis complete, so the failure can be prevented, ranks somewhere between 1 and 5 on our scales. The best after-the-fact inspection never scores better than a 5 for detection.

    So for our toast example, if the setting was based on a color selection by the user, and there was technology in place to adjust the heat level or timing if the color of the toast was too dark, that would be a 1. If the setting required validation from a user if they selected “burnt toast” color, it would be higher, but under 5, since there is still an opportunity to fix the toast before it cooks. If the cook checks the toast level every 30 seconds, it couldn’t score better than a 6, if we waited for the customer to refuse the toast it would be a 10.



    What do you use occurance for?

    I would say that prevention would effect occurance score and not the detection as you are reducing the possibility of it happening.

    I see these as two distinct elements and the poor score would indicate which area I needed to work on.

    Which one you would work on first would be depend on the process in most cases prevention will be the most cost effective unless the total cost of prevention is higher than scrap and or rework.



    As I mentioned, pure prevention controls could be applied to occurrence if the nature of the control is such that the cause could not occur. However, if the cause has occurred, controls vary in their ability to prevent the effect of the failure mode from reaching the customer.

    The toast example isn’t the best to illustrate this. Let me use an example from a chemical process.

    Failure mode: insufficient catalyst for reaction
    Effect: product strength does not meet spec
    Cause: plugging in catalyst feed line

    I would consider a control that monitored the catalyst flow rate or the catalyst feed valve position to be more preventative and have a better detection score than a control that periodically sampled the reactor contents checking for the right catalyst level and that having a better detection score than a control that samples and analyzes the final product strength.

Viewing 9 posts - 1 through 9 (of 9 total)

You must be logged in to reply to this topic.