Lab-based Quality Control means the production line makes product, a sample is taken to the lab, one or more tests are done, some tests may take hours, days or even a week, the results are reviewed and then materials or process conditions are adjusted to keep product on-spec. Now 6-Sigma has helped keep product characteristics within spec by removing variance in materials and the process, but the key word is "helped" and life can be a bit complicated.
Our Chief Technology Officer was a process engineer at a 3M plant where they made photo paper coated with silver-based complex solutions with many odd mixes of different chemicals applied in multiple layers that chemically interacted with each other. The product was expensive and each "jumbo" roll of finished photo paper coming off the line was worth about US$100,000 in today's dollars. Here's the rub... Let's say you make a "jumbo" of materials, $100,000 committed. A sample is taken. Tests are done while the next "jumbo" is being made. The test result would arrive about the time the 2nd "jumbo" is complete; $200,000 committed. The third "jumbo" is started about the time it is discovered the product is "bad". An adjustment is made to the materials or the process conditions in the HOPES that it will solve the quality problem. The test from the 2nd "jumbo" arrives and it is bad of course, because no adjustment had been made to it. Now, somewhere within the third "jumbo" the change takes effect. Even if the test is good, 2.5 "jumbos" have been made and $250,000 is destined for "the dumpster" (recycling) and a lot of testing later has to be done to try to find where the change occurred within it. That is the BEST case... But did the adjustment help? You won't know until the third "jumbo" completes. If it remains out of spec so much money has been scrapped that the line has to shut down and the Plant Manager called, perhaps at 3 AM, informing them you just made a mountain of very expensive crap.
Product characteristics are substantially based upon the materials characteristics and the process conditions used to create the product. If you have a model or models that relate materials and process to quality you could supply the current materials and process data to the model and have an estimate of the quality, immediately. Putting such a model on-line would give a real-time indication of quality immediately as product is made, not hours, days or even a week later. As you change setpoints or materials you will see the quality estimate respond in real-time. This is a major benefit.
The next "by-product" of such models is being able to query them about the effects materials and process conditions have on quality. This includes "key drivers" identification (which materials and process conditions have the greatest effect on each quality metric) and 2D and 3D response curves and surfaces. This substantially increases product understanding.
Certain factors are in your control while making a product. Other factors are already "cast in stone" and are unchangeable. If we determine which can yet be changed to affect the product's physical characteristics we mark those as "controllable". We then use a model-based optimizer that takes the uncontrollable input's values, puts them into the model and then searches the combinations of values of the "controllable" variables in order to target one or more quality metrics. The resulting suggested "controllable" variables' values are offered as recommendations to target quality to specification. This works well, even with many "controllable" variables. We've done 60 of them at one time, though normal situations usually involve a handful at a time. The operator can take these suggested setpoints as an advisory, hence we call this open-loop advisory control.
If the advisories in the prior step are good and are trustworthy, they can be written to the control system as recommended setpoints. The control system is configured to screen the values to make sure they are reasonable, then transfer them into registers as actual setpoints. This is often done in a "cascade control" scheme. With this, you have a setpoint for quality across one or more product characteristics and the system sets the control system in real-time.
Since we have an ongoing stream of materials properties, process conditions and lab tests (yes, lab testing continues) our system can access that data automatically and compare its estimates with lab tests and self-calibrate accordingly. No work on your part. We can monitor the solution remotely as we do world-wide today for other customers.
Benefits achieved include:
Lab-based quality control can be too slow, creating unnecessary waste
The solution enables you to see quality in real-time.
The solution provides substantial improvements in product, process and materials understanding.
You can receive recommended setpoints in times of need, based on current conditions at the moment.
You can close the loop on quality
The system can be self-maintaining.