shallowmike said:
We have a process where we weld contacts using a resistant weld process. The welds are checked through a sampling rate at determined times by a destructive test. The contacts are welded to a backing material in strip form, exiting the welder and then processed into piece parts through a press. We can't check every weld, as you can imagine we would have no end product. There is no way we can run SPC on the process to give us a validated detection tool. This leaves us with sampling rates and our experience that our processes are robust with not many defects produced. It does however to a degree show the process is out of control. When doing the numbers a high RPN is generated. Because our process is robust would that in itself qualify as a reasonable detection? With regards to severity if a contact becomes detached in the end customer product who determines the severity number, as we are third tier suppliers do we liase with our customer with regards to this number. We have data to support our occurence numbers its the other two factors that are causing discussion. How would an auditor view this? I take your point about not using a trigger number as this could have an influence when doing the
fmea. Would an auditor want to see some factual support to assess how you arrived at the number where we would deem necessary to start implementing corrective action. Auditors dont like low or high numbers. What do you think?
When you recommend actions--and why--in the
PFMEA process is a function of experience and common sense. As I suggested earlier, there is no definite connection between RPN values and whether or not something can be prudently improved. I commend you for your desire to have data to support your decisions. When it comes to the Detection value, the criterion is the likelihood that the defect (or defective) in question will be detected A) before value is added, or B) before it is shipped to the customer. You should have historical data to indicate how likely it is that bad product will be shipped, and how often bad material is detected before subsequent operations. From you can estimate the likelihood, and then rate it on the FMEA 1-10 scale.
Severity is another story, and is not without controversy. IMO, job shops
should not attempt to divine how a given defect will affect end use of the product. That information should be incorporated in the specifications. You should base the severity rating on how a given defect might affect
your operations. If you make the part in accordance with the specifications and it doesn't work in the intended application, you have made a quality part, so it makes sense to let the customer worry about end use and for you to be most concerned with how nonconforming conditions will affect
you.