I was teaching about Statistical Process Control (SPC) and control charts today and a student asked an interesting question: Are the upper and lower control limit the same as the design tolerances on a manufactured part?
The answer is that they are not the same: the upper and lower control limit are based on the statistical information obtained from observing the process, while the design tolerances are what the designer wants to achieve.
This leads to the next question: what is the relationship between tolerances and upper and lower control limits? This hits on the core philosophy behind SPC. Traditionally people like to inspect quality into a part. To do this, every piece must be inspected and the rejects discarded - there is no thought put into the process that created these rejects. It can be a lot of cost and time to inspecting every part. In some ways it is easy to image every manufactured part being inspected, but what about services? Quality is also important in service. How can each service be inspected before it is delivered? Lets say the service is a bank teller. It is completely unfeasible for a third person to inspect every thing the teller is doing before the service is actually rendered to the customer.
This is where SPC comes in. SPC is not about inspecting every single part to see whether or not they are in tolerance, rather SPC is about monitoring the process and improving the process so that the process will automatically provide the desired quality. In other words, with SPC the process can be improved so that the upper and lower control limits fall within the desired tolerance.
This is a radically different way to look at manufacturing and servicing processes. With the old way of inspecting every part, people continually work in fear about ever making a mistake. The fear of an error is the predominate thought on their mind.
With an SPC controlled system, the focus is on trying to understand the process, where there is potential errors or defect (deviations from the desired outcome), and improving the process so the desired outcome can be achieved more predicatably.
Friday, October 16, 2009
In a regular annotation note, create a stack that looks the way you like. It will take a couple of trial and error attempts since there'...