Now viewing: statistical processing

Ken Roney
By Ken Roney
Tuesday, September 10, 2013 - 11:34

In part two of this series we discussed Process Control and, specifically, what “control” means. The control chart is very helpful tool in identifying unwanted variation in a production process. The goal of statistical control is to identify process variation and to determine which variation is beyond our control, that is, variation that is inherent to the particular process we are employing, and which variation is “special” or assignable. This type of variation is outside of the “normal” variation we typically see in a process. It can be identified using statistical tools and reduced/eliminated from the normal process.

First, let’s discuss variation. In a perfect world there would be no variation. Whenever we turned on a machine it would turn out the exact same part every time we used it. As you are well aware, we don’t live in this perfect world. There are many inputs to a given production process that must all work together to produce parts that will satisfy customer requirements. Machines can fail, internal parts wear out, temperatures change, molding tools wear over time, materials can subtly change from lot to lot, etc. By measuring key characteristics of the finished part we can evaluate the process that made it. Using control charts we can identify “normal” variation from special cause variation (a mold wear issue or machine settings that are not set properly) and take steps to eliminate these special causes and return the process to “normal”.

Two charts regularly used at Elite Plastics are Individual Charts and Xbar-R charts. Both charts feature a Center Line which is the mean (average) for the particular data set being studied and upper and lower Control Limits, which are statistically calculated from the same data set. Measurement data for key characteristics are collected and entered at regular intervals over time.

This example shows the center line and upper and lower control limits of a contr

This example shows the center line and upper and lower control limits of a control chart.

Note that specification limits are not used for these charts. Control charts show process variation. It is up to the company to adjust the process to match the specifications required by the customer.

The control limits, calculated statistically, represent plus/minus three standard deviations from the average (mean). These lines represent the threshold at which the process output is considered to be statistically “unlikely”. In other words, the control limits represent the division of the “natural variation” of a process from the “unlikely” variation that is occurring due to special or assignable causes that can be eliminated from the process. Data points that appear outside of these control limits or unusual data runs (data increasing or decreasing over 6 or more successive measurements) should be investigated so that the source of the assignable variation can be determined and eliminated. An advantage of identifying and dealing with this variation early is that steps can be taken to prevent problems before any out-of-specification parts are produced. This maximizes machine and material efficiency which, in turn, lowers production costs.

This concludes our three part series on Statistical Process Control at Elite Plastics. I hope that this has been helpful in describing our efforts to maximize our production efficiency by building quality into our production processes. 



By Ken Roney
Tuesday, August 20, 2013 - 00:00

An increasing number of our customers at Elite Plastics require that we develop, record, and make available upon request, process capability data for key characteristics found on the print. Process capability indexes (most commonly Cpk) are an indicator of how a production process is performed with regards to key process control dimensions and their relation to customer specifications. Besides the obvious reason of satisfying customer requirements there is another reason to develop, and use, capability indexes. They can be used as predictors of future performance and are a good way of measuring the ability of our production process to produce good parts consistently.

Accuracy and Precision

The letter “C” in SPC stands for “control,” but what are we controlling? Statistical analysis of a production process is ultimately all about reducing variation in that process and to identify, and separate, the natural causes of variation that are inherent in all processes, from the special, or assignable, causes that can be controlled, adjusted and/or eliminated. 

The accuracy and precision of a set of measurements can be illustrated as follows:

The goal is to confirm the consistency of the process. With that in mind, it is extremely important that the company has a good handle on the inputs and variables of the given process and have the ability to make sound adjustments at the qualification stage to the inputs that are within their control. Process capability is also about both accuracy and precision. 

A Cpk index takes into account both the accuracy of the measurement and the precision of the measurements around the average. Statistical software takes measurement data and shows the process capability as a single number that represents the process’ ability to provide both accurate and precise product. For most of our customers we aim for a process capability index (Cpk) of 1.33 or greater.

Process Capability is the ability of the process to produce product that is both accurate and precise, that it consistently produces product that meets customer requirements with a minimum of variation over time. The benefits of creating robust processes that deliver a high level capability are obvious – it means a maximization of production time and materials, reduced scrap and/or rework, and the ability of the production system to deliver more products in the minimum amount of time possible.

Stay tuned for an upcoming blog discussing control charts and their use in our production processes.