Wow, this thread has a lot of retro ideas - how far we have gone in just a few years!
1) First step: gage calibration and gage R&R, make sure you are using the correct gage for the job!
2) Incoming inspection: sure, everybody wants to believe their vendor sent the right stuff, but it sure doesn't hurt to make sure. Sometimes the first metal forming step in a process is done on the truck from the steel warehouse to your dock! Am I right? Of course I am. Check the chemistries on the certs - I have seen bad material certified bad (or good for that matter), and they missed it and shipped it. Cheaper to check it before screwing up your scheduling with scrap and rework.
3) Screw machine/CNC/grinding (and all other precision machining) inspection: Check the tightest dimension and/or the dimension that wears the quickest for each tool that makes a final dimension off of the machine. Keep the data separate by spindle (especially when doing capability verification). SPC using X hi-lo/R chart if you wish (See
Statistical process control for precision machining), or at least run to 75% of the specification. Do not use Xbar-R charts, it is the worst and most useless chart for machining,
bar none. Do not use precontrol charts, they also provide the wrong direction - they encourage overcontrol. Do not run to the mean. That is back yard hack. For ODs: set up near the lower control limit, and let the tool wear to the upper control limit before removing for replacement or sharpening. For IDs, start at the upper limit and let the tool wear to the lower control limit before removing for replacement or sharpening.
To determine the frequency, do an X hi/lo-R chart capability study of 100 consecutive parts. Do each spindle separately. Calculate the linear regression on the upper and lower X plots. Extrapolate them to the control limits, if 100 parts does not provide enough wear to make it to the control limits on their own. For the number of parts is takes to go from where the X-lo is at the lower control limit to where the X-hi hits the upper control limit, divide that number of parts by 5. That is your minimum frequency for charting.
Just because you chart, and the data is in control, can you assume the product is ok automatically. You should inspect the last part (from each spindle) prior to moving the lot to the next operation - or a special cause (broken tool, etc.) might have occurred, you would not know it, and you are moving
scrap. ALWAYS check the last parts to ensure the tools did not break or chip out. Otherwise, you will be sorting all over the shop, rather than containing at the operation.
4)
Do incoming inspection on plating thickness and hardness, unless you have some rare world class suppliers. Train your outsourced precision machining suppliers (grinders, etc.) on
correct SPC described above, and then sample incoming product. Do not look for Cpk of incoming product, if they are running their process correctly, it will not be relevant, and your sampling error will give you ridiculous conclusions.
Statistical process control for precision machining - Sampling Error Just be sure that all of the sample you take are in spec.
5) If you are doing heat treating and plating, you need lot traceability. If you end up with a raw material that does not harden correctly, you do not want to throw out any more product than the lot of material made. Think - what is the point of traceability? It is not just extra paperwork. Somebody got burned - bad - and they figured out traceability is the only way to keep it happening to you. Learn from other's mistakes. Some people make lots by operator or shift...just in case. Really helps during training a new employee.
There - now we have freshened this up a bit!