You say you predict a part will fail. Are you a medium? Presumably not. Do you predict that good parts, within their MTBF, will fail? I don’t think so. So your prediction has to be formed on the basis of some observed defect. So you take corrective action and remove that defect. It is corrective, not preventive.
Ummm, well, sorta. Situation: Current production. We are in a 'team' meeting of production staff. We review certain process data and see some patterns. We see no evidence of a 'defect', only data which we review every meeting. We decide that it is evident first shift production is not as efficient as second shift. Here, with NO DEFECT PRESENT, we PREDICT there is a difference which, if we can identify cause, may allow us to improve first shift through-put. We have acted on other issues and decide to take on this issue. We form a team, investigate, find th cause and act. We do not have to be 'mediums' to predict. In fact, we all do it every day.
4.14.3 b cites measureables and says you must look at this data and analyze it. Note that it cites "...appropriate sources of information such as processes and work operations..." which is current production. This means you must make PREDICTIONS based upon the data at hand. There does not at all have to be a specific problem (how do we define PROBLEM?) or defect to make a prediction and there is no bearing on the planning stage - we are in production and have been for a year.
The latest ISO9001:2000 draft is reworded and 8.4.2 starts out to say:
"The organization shall establish a process for eliminating the causes of potential nonconformities to prevent occurrence."
I take "...eliminating the causes of potential nonconformities to prevent occurrence." to indicate that no problem or defect has yet to occurred.
Kim, You say that preventive action prevents an occurance. But we are not talking about putting out fires.
Welllll - you are mixing this up, maybe. There is no fire yet which is why it appears to be preventive (we all know about oily rags and spontaneous combustion). The problem is in part due to the example. If you spot oily rags and it is a 'problem' (safety issue), it is akin to a defect and your reaction will be corrective.
I stick with my earlier 'definition':
Corrective Action is a Response to an observed problem / defect.
Preventive Action is Predictive (a problem *may* occur based upon analysis of data).
You say:
So, if all dealing with nonconformities is corrective, you can only take preventive action before you find nonconformities. Therefore preventive action involves detection.
Nope - I disagree. Preventive action does not involve detection. To detect something there has to be something (a defect?) to detect. If it hasn't happened yet, you can't detect it. This is precisely why I say it is predictive in nature.
And Preventive Action is a big part of Continuous Improvement. I also see Preventive Action as a current production issue unrelated to the planning / design phase.
Last - if you want to say that in preventive action you are REACTING, it is a reaction to data. I see this data. It tells me a problem might occur (I am PREDICTING as no problem has yet actually occurred - I just believe, based upon my background and experience, that there is a 'hole in the system' so to speak). In this sense, yes - both are reactions. But you are also reacting to data and experiences in the
FMEA stage. So - "How do you define the word 'reaction'?" and when do you use it?
Nuff said here - I'm in a hotel in NY and I gotta go beddy bye. Long day tomorrow.