Sie sind auf Seite 1von 2


CHARTS Median charts are another alternative to the Xbar and R

charts for control of processes with measured data; they yield similar conclusions
but have several advantages: 1. Median charts are easy to use, and do not
require day-to-day calculations. This can increase shop floor
acceptance of the control chart approach. 2. S ince individual values (as well
as medians) are plotted, the median chart shows the spread of the
process output and gives an ongoing picture of the process variation. 3.
S ince a single chart shows both the median and spread, it can be used to
compare the output of several processes, or of the same process at
successive stages. CONS TRUCTION OF THE CHART The method of constructing the
median chart is similar to that of the Xbar and R chart. Initial data is collected
in the same manner as for the Xbar and R chart. Generally the median chart uses a
subgroup sample size of 10 or less; odd sample sizes are most convenient. S ince
only a single graph is plotted, the scale should be set to include the larger of
(a) the product specification tolerance plus an allowance for out-of-specification
readings, or (b) 1-1/2 to 2 times the difference between the highest and lowest
individual measurements. Also, it is recommended that the gage used should divide
the product tolerance into at least 20 increments -- the graph scale should agree
with the gage. Each individual measurement for a subgroup is plotted in a
verticle line for each subgroup. The median for each subgroup is circled (the
middle value; if the sample size is an even number, the median will be mid way
between the inner points). To aid in interpretation, the subgroup medians are
connected with a line between them. Each subgroup median (X) and range (R) are
entered in the data table. Note that the symbol for median differs from the symbol
for the mean in that the X has a wavy line over it instead of a straight line. This
wavy line is called a "tilde". Therefore we refer to the median as X-tilde and to
the mean as X-bar. CALCULATING THE CONTROL LIMITS There are two variations
of control limit calculations for median charts. The difference rests in which
value we use for the central line. We may use the median-median value (Xdouble
tilde) or the mean-median value (Xbar tilde). Either may be used. However, the use
of Xdouble tilde requires that all Xtilde values be ranked in order so that the
middle value can be found. The Xbar tilde is found by adding all of the Xtildes and
dividing by the number of samples [(X(1) +X(2) +X(3) + . . .+X(n))/k]. The value of
Rbar is found in the same manner as for the Xbar and R charts.
************************************* NOTE TILDES AND BARS WILL NEED TO BE DRAWN ON
THE FOLLOWING ************************************** Formulas for calculating
the control limits are as follows: UCL(R) = D(4) Rbar
LCL(R) = D(3) Rbar UCL(x) = X + A(3) Rbar LCL(x) = X -
A(3) Rbar UCL(x) = X + A(5) Rbar LCL(x) = X - A(5) Rbar
Where D(4), D(3), A(3), and A(5) are constants varying by sample size, with the
values for sample sizes 2 to 10 shown in the following table. n 2 3 4
5 6 7 8 9 10 D(4) 3.27 2.57 2.28 2.11 2.00 1.92 1.86
1.82 1.78 D(3) 0 0 0 0 0 .08 .14 .18 .22 A(3) 1.88
1.19 .80 .69 .55 .51 .43 .41 .36 A(5) 2.22 1.26 .83 .71 .56
.52 .44 .42 .37 After calculation the limits are placed on the chart. The
median chart may be interpreted the same as the Xbar and R chart. S ince there is no
difference between the two charts in terms of calculation of the Rbar, it is
possible to obtain the standard deviation of the individual values in the same
manner as was used for the Xbar and R data [S D(X') = Rbar/d(2)]. Provided the
process has a normal distribution, the mean and median will both be very nearly the
same. As long as the medians and ranges are in statistical control we can use the
estimate of the standard deviation of individuals to calculate the expected limits
for individuals [X +/- 3S D(x')]. ONGOING PROCES S CONTROL For ongoing
process control where control limits are based on prior data, the charting process
can be simplified considerably. It is possible to use a single chart with the
scales set at the same increments as the measuring device and with the central line
and control limits for medians already entered. S ince the actual measurements
are recorded on the median chart in order to obtain the median value for a given
sample, the range is automatically plotted on the median chart. It is not necessary
to plot a separate range chart. Instead, a card is prepared with the range limits
marked on the edge. The range for a sample can then be compared to that marked on
the card. Any sample range beyond the limits on the card is considered out of
control and is enclosed in a narrow verticle box. Of course if any plot for
range or median falls outside the limits, the operator must take appropriate action
to correct the process. It must emphasized that only the circled medians are to
considered when assessing the control chart. S ome plots of the
actual values may fall outside the control limits for medians.
S ince the control limits are based on the medians,ONLY MEDIANS are to be used in
decisions for action. The fact that an actual value may plot outside the control
limits on the median chart does not mean that the process is out of control. Of
course the actual plots are used in decisions relative to control of ranges.