In Part Two of this blog I discussed: 1) the cause of defects, 2) statistical analysis techniques, and 3) different statistical analysis tools that are available to youclick here to refresh your memory by reviewing Part Two of our blog
How do you achieve consistent quality? By having very well-defined processes that allow little or no variation, which can be benchmarked and measured. This type of benchmarking and measuring is what statistical analysis provides.
So, how can you effectively determine if your process is predictable or even capable of performing the task of delivering to the customer’s expectation? This is done by Statistical Process Control (SPC).
SPC was pioneered by Walter A. Shewhart at Bell Laboratories in the early 1920s. Shewhart developed the control chart in 1924 and the concept of a state of statistical control. It is used in the Control phase of the DMAIC methodology for measuring and determining control limits. It uses control charts to determine whether a process is stable and capable of meeting specifications limits.
This is a simple representation of a run chart. Here is some measured output that changes over time. Imposed boundaries or specifications are generally outlined by customers. We can plot these on a run chart to indicate how this process is performing relative to the specifications.
A practical example of this is the manufacturing of widgets (nuts and bolts), both nuts and bolts have some specifications, the nuts have to be certain diameter, thread size and length to be able to fit onto the bolts. In a machine shop the manufacturing of these nuts and bolts are done by machines that are spitting out thousands of these nuts and bolts hourly, on different machines, run by different operators, and done in different shifts. Like all machines, they are subject to breakage, need maintenance, get out of spec, or the operator gets tired or is not feeling well on one day. Over time these conditions affect the process. With sampling we can generate data on the thickness at different intervals. We can then take that data and plot a run chart to determine whether the process is consistently meeting customer specifications.
This is a much better way of determining whether it is not, versus hearing about it from a very dissatisfied customer who is receiving defective products.
Plotting this run chart on an extended basis and inserting the control limits – such as the center line, the upper and lower control limits, and plotting the data points – allows us to analyze other things, such as, do we have points that are unstable, as indicated by the point that is higher than the upper control limit. Is there instability due to trends, which are determined by a number of consecutive points in an upward or downward direction along with several other tests that were specified by Shewhart as determinants of an unstable process? This is the power of control charts and Statistical Process Control. It allows you to determine things that you would not be able to do using non-statistical methods.
Shifts are just another indication of an unstable process. It simply means that at some point the process is humming along nicely with ebbs and flows and some indication of a centerline. Suddenly, the centerline shifts to a higher point and the process is now operating at a different level of operation (good or bad) this is unstable and unpredictable as to what may happen at any given time.
Process Capability analysis requires that you have a stable process. We can use a control chart to ensure process stability and specific criteria to determine whether the process is capable and acceptable. If the processes have varying levels of dispersion (fatness relative to the center) and are shifted to one or the other side of the specification limits, they are not stable. If the process over time stays centered around the specification limits and tight in terms of dispersion, it is stable. When the process stays within the boundaries of the specification limits, it is determined to be capable of meeting the customer’s expectations.
Generally, if the process spread is les specification spread, the process variation is potentially capable of fitting. In the example above, the process is not capable, because the process is too wide for the specifications. This is analogous to trying to fit a large SUV type, like a Hummer or a monster truck, into a one-car garage. No matter how you try, it won’t fit – it is not capable of fitting the specification of the garage.
There are parameters and ways we can determine whether a process is capable. The ultimate goal should be to fit at least 3-Sigmas between the center of the process and the nearest spec limit, which would in equivalent to a 6-Sigma process and result in having only 3.4 defects per million opportunities.
What can this mean for your company? Well, using these very powerful tools and with an effective process methodology, you can drive process improvement through your organization by understanding the voice of the process and ensuring that the process is stable and capable. You can institute continuous improvement by creating a control plan that applies Statistical Process Control methods. You can realize better business performance and achieve improved results driven by a proven methodology, tools, and measures.
These strategies can be applied to any industry sector and to any company involved in manufacturing or providing services.