There are many misconceptions around statistical process control. It is seldom applied in China, yet it can help control processes and ensure a consistent output.
To understand how this tool works, the best is to understand the logical steps to put it in place in a factory.
For example, if the metal hardness is lower than expected on a screwdriver, that product might not be able to perform its intended task. Hardness is a CTQ characteristic.
What qualifies a process as “critical”? If something goes wrong with that process, it will probably have a sizeable impact on at least one CTQ characteristic.
Critical processes are often indicated with a * on the control plan.
Many modern pieces of equipment collect and analyze data, and then issue an alert when they get out of control. This is very uncommon in China, so let’s assume this is not the case; it means you will need to go to steps 4 and 5…
For example, the variables that might impact the output of a gluing process are listed below:
Based on this, you might conclude that the viscosity of the glue and the ambient humidity are two variables that need controlling.
How to get that information? It might come from engineers, chemists, physicians, etc. but also from the operators that have been working on that process for some time and may have noticed some cause-and-effect relationships.
For more details, I suggest you read ‘Reducing Production Costs’ by Don Wheeler.
Gather data regularly about those variables – for example 5 random samples every 4 hours.
Teach the production operators and leaders how to make basic calculations and plot them on a chart – for example, the most widespread tool is the classic “Xbar – R” chart. For simplicity, we tend to set a goal for the Cpk index value, and let operators plot the evolution of that index over time.
Now you have made variation visible. The next step is to find ways to reduce it. If you use control charts, ideally it looks somewhat like this:
As long as the Cpk index is below target, engineers and production leaders are encouraged to test different approaches. A Cpk target of 1.0 is often attainable within a few weeks. 1.33 is more challenging. 1.66 is much, much harder!
If necessary, use another statistical technique called DoE (Design of Experiments). It can help you get close to the optimal values for the variables that impact your process output. Be prepared – there are several available approaches here and it can get fairly complex.
Let’s say you manage to reduce variation. Set a new target, and keep controlling the key variables. If your process characteristics drift in one way or another, your statistical process control system will alert you. Otherwise, don’t temper with the process. Don’t do constant little adjustments – they will mechanically make the process “unstable” in statistical terms.
Have you seen SPC at work in China? If so, in what sectors? It seems to be quite rare except in a few industries such as semiconductors.