Introduction
Control charts are a very important tool in quality
control. We say “A picture is worth a
thousand words” because images are better at conveying complex ideas and most
people find it easier to understand large amounts of data visually than in
tabular form. Variability in a
manufactured product is unavoidable, but it is important to understand if the
variability is controlled variation or uncontrolled variation. Controlled variation is attributed to chance
or common causes and is a consistent pattern of variation over time. Uncontrolled variation is attributed to
assignable or special causes and is an inconsistent pattern of variation over
time. The primary use of control charts
is to quickly identify uncontrolled variation attributable to special
causes. This can be difficult depending
on the type of special cause. Some
special causes, such as crusher liner wear, may cause a drift or a slow change
to the gradation. Other special causes,
such as a change in mix design materials or equipment malfunction, may cause a
step-change that is more abrupt (Obla, 2014).
In its most basic form, a control chart can be any chart
containing control limits. However, the
term “Control Chart” is typically attributed to Dr. Walter A. Shewhart and is
also referred to as Shewhart Charts or Process-Behavior Charts. Shewhart charts are used to monitor the
process average and process variability using either standard deviation or
range. Control limits are typically
upper and lower boundaries outside of which the process is statistically
unlikely to produce results (they represent the limits of the natural process
or controlled variation). In other
words, control limits should be computed from the process rather than be
specified limits, such as specifications.
If your control limits are outside of your specification limits then
your process is telling you that it is statistically likely to produce
failures. Control charts are used to
identify signals such as when a data point is outside of the control limits, or
two out of three successive points fall on the same side of the average (NSSGA,
2013). Stonemont Software includes rule
checking to help identify common signals.
differentiate between particular methods of
analysis. These charts are very powerful
tools that can be used to analyze gradation results as % passing, % retained,
or % individual retained; all quality test results, such as FM, AC Content,
Specific Gravity, Slump, and Unit Weight/Density; process results such as
crusher amps, liner hours; both overall and within batch concrete compressive
strength results; and concrete batch results.
Run Charts
Run charts are probably the most commonly used chart in
Stonemont Software because they are the most easily understood. Run charts show individual data points and
the average value with control limits that are +/- 2 x Standard Deviation
(SD). When data are normally
distributed, control limits at 2 x SD should contain about 95.4% of the data
points given a data set of adequate size.
Run charts help monitor the process mean and are useful for
understanding the relationship between control limits based on 2xSD and
specification limits, identifying trends using a best-fit trend line, and
changes in mean using a moving average line.
Control Charts
In Stonemont Software, our standard control charts are based
on Shewhart charts and they provide the ability analyze data as individual
charts, average and range charts, and moving average and range charts. Average and moving average charts use subgroups
of data points, which have the effect of smoothing out some of the erratic
individual data points making overall trends or changes in mean easier to
identify. However, individual charts are
more suited for data that is collected periodically to ensure that a change in
the process is not missed (NSSGA, 2013).
Stonemont Software includes a special control chart for
compressive strength data that shows the within-batch variation or the
variation between cylinders made from the same concrete sample. This chart also shows the average, moving
average, limits, design and required strengths.
The range chart can be the actual within batch range, the normalized
within batch range expressed as a percent, or the within batch coefficient of
variation (CV), which is used as a measure of testing variability. The user can define a limit when using the
range % or CV chart.
Cusum Charts
Cusum charts are designed to quickly identify a change in
the process mean or average; commonly a 1 x SD change in the process mean. Cusum charts can be represented either as a
standard or tabular form of the cusum. Stonemont
Software uses the tabular cusum on individual cusum charts and the standard
form on multi-variable cusum charts. The
tabular cusum chart is much easier to read and understand because control
limits can be used to quickly identify a change in mean. However, a multi-variable cusum chart is
useful for identifying correlations of change between variables such as whether
concrete compressive strength changes correlate to those of concrete density,
slump, or temperature. The
multi-variable cusum chart shows how changes in the batching
process on 6/10/2014 led to an increase in slump and corresponding decreases in
unit weight and 7 and 28 day strength results.
Based on these correlations it appears that the cause may be an increase in
water added to each load.
Summary
Control charts are an extremely useful tool in quality
control. Stonemont Software provides
several different control charts, which allows quality control personnel to use
the control chart that they feel is best suited for their application. Not only can these control charts be manually
generated from within the software, many of these charts can be automatically
generated and delivered via email for multiple variables such as the run chart,
control chart, and cusum chart auto-reports (auto-reports are available in our
enterprise and hosted/cloud editions). Automated
reporting is critical for your most common products.
References
Obla, K.H., 2014, Improving Concrete Quality, CRC Press.
NSSGA, 2013, The Aggregates Handbook Second Edition, NSSGA.
Adrian Field
Michael Rodriguez