ARTICLES

Stage I and II SPC

The BIS.Net Team BIS.Net Team

Many Quality Professionals apply SPC software incorrectly according to some experts.

Shewhart, the recognized founder of statistical process control specified two stages for SPC. The first stage is about the initial establishment of the process. Control charts are foremost used in this process to remove unnecessary variation by identifying and removing assignable causes of variation. Points falling outside control limits are deemed as being due to assignable causes. At this stage limits are placed around the process average, estimated from the average of all samples, or the known process average. Once the process is ready and assignable causes have been removed the second phase can commence. Here it is common to use fixed control limits determined after assignable cause variation has been removed and the process is predictable. The best way of determining fixed limits after stabilizing the process is to enter the process target into the Setup and then determine control limits around this target by running the analysis. These should then become the fixed limits entered into BISNET SPC. From then on Stage II SPC is applied with fixed limits. The charts can then be used to ensure the process remains on target and detect occurrence of new assignable cause variation.

X-bar/ Range, X-bar/Sd and Individuals/ Moving range charts are best used for Phase I and can also be used in Phase II with fixed limits as explained above.

During Phase II Moving Average and EWMA charts can be used to detect changes in the process mean faster than X-bar/ I charts. EWMA charts are the preferred option over moving average charts. However, X-bar and Individuals charts should still be used to detect assignable causes other than those that cause mean changes.

To most rapidly detect changes the Cusum Chart should be used during phase II.

The Moving Average, EWMA and Cusum charts are not appropriate for Phase I as they are designed only to detect changes in the mean, not outliers caused by assignable causes.

The EWMA does not have fixed control limits as these evolve continuously. Instead the center line is fixed for Phase II. Similarly, the Cusum chart does not have fixed limits. Limits are dependent on the Process Sd and amount of concern and average in-control run length, which in BISNeT SPC is factory set at 370. The moving average fixed limits apply only after more than the number of points equal to the period have been plotted.

Once the process is in control and capability with a CP index of at least 2 has been attained modified control charts can be applied as an option during Phase II. These allow process drift. The recommended option is the six-sigma process modified control chart

Zone charts are provided instead of supplementing standard control charts for both Phase I and II. The practice of supplementing SPC charts with run tests is not an option with BISNet SPC as it only increases false alarm rates.

Both classical and machine powered versions are provided. The classical version assumes normality. Contrary to general opinion and even Shewhart’s conclusion, classical SPC is not as robust to non-normality as believed. If there are too many out-of-control limits that cannot be identified due to assignable causes, then chances are that data is non-normal. In this instance use the machine powered charts. The machine powered charts can be used all the time, because they work equally well on normal data and provide statistically the same results when normal.

Machine powered charts are typically used during Phase I to stabilize the process and obtain fixed limits for Phase II. Once fixed limits are used neither classical or machine optimized technology is used because the limits have already been determined. To use fixed limits, select from Classical Charts, even though classical charts technology is not used. BISNet Analyst disables the use of fixed limits for distribution optimized charts.

Within subgroup variation is used for classical SPC charts. This has led to the practice of removing the effect of assignable cause variation from control chart calculations by removing data from out-of-control points. Although BIS.Net Analyst, in keeping with current practices, provides this feature, it is not recommended as this only removes the out-of-control point data. In practice even in-control data may have been affected by the assignable cause. It is better to remove the assignable cause and repeat the analysis with new data to obtain more reliable control limits.

Distribution optimized charts are used to obtain reliable control limits when the data is non-normal. It is possible that assignable causes, especially if there are many out-of-control points, will affect the non-normality analysis. It is recommended that before applying the algorithms a change analysis is performed with the change analysis module. This will identify changes in the process mean and outliers due to assignable causes of variation. If present the process needs to be first stabilized before obtaining fixed control limits with distribution optimized technology. (Change Analysis is our recommended option as a more modern alternative to SPC charts. The choice is yours.)

Distribution optimized Attributes charts are offered as an option. They are optimized in the sense that exact limits are used, as recommended by some scholars. Exact is in the sense that the probability points are the same as the standard variables control charts based on plus or minus 3 standard error. Although the standard approach uses plus or minus 3 standard deviations, the end point probabilities differ because the distributions are not necessarily normal. Our research has however indicated that similar results are obtained with both technologies.

Download the Inferences APP, comprised of mainstream and machine-powered analytics for statistical analysis

Analytics as a Service (AaaS) for Quality

Drive quality improvement through actionable insights using analytics you can trust! Use up to 200 analytics tools downloadable through a suite of Apps!

FREE usage of the analytics Apps for quality improvement
  • Augmented with machine-powered smarts
  • Always updated with the latest tools and features
  • No licencing or fixed subscriptions - Pay ONLY for the analysis you run from 20 USD cents per analysis, billed monthly! Set a budget so you don't exceed!