You are on page 1of 3

UNIT VI Xbar AND R CHART INTERPRETATION In the last section we calculated Xbar's

and R's, plotted the values on the chart, calculated center lines and control
limits, and drew these on the chart. The chart is complete from the standpoint of
construction, but it is still essentially useless until we are able to interpret
what it can tell us about the process. THE CENTER LINE We have learned that the
range is a measure of the variation among the measurements making up the sample. If
all the parts were equal for the precision used, then all of the measurements would
be the same. A range of zero might be desirable but it is not necessarily
economical; it is a quality ideal. Since the center line on the R Chart represents
the average range, it is the one value which best represents the R-values. We
desire that the value of R be small but we do not know how small it should be to be
satisfactory. The center line for the Xbar chart represents the grand average
(Xdouble bar) which is actually the average of all the dimensions recorded. We
found that Xdouble bar was +0.66 in our example. Since the increment of measurement
was 0.0001", our average dimension was 0.000 066" (+0.66 x 0.0001") larger than the
nominal dimension of 2.1815". Therefore the true process average was 2.181 566".
This means that the process average was about 1/10,000 of an inch larger than the
intended value. This may seem a small amount but since the specification was only
+/-0.0005, we are not yet in a position to judge whether this is an important
difference or not. We will discuss these problems later. THE CONTROL LIMIT FUNCTION
It was pointed out earlier that variation is a fact of life and that a predictable
pattern of variation exists as long as the process is operating under a stable
system (the 4-M's). The plottings we made on our Xbar and R Chart show a certain
amount of variation. We may ask ourselves whether this variation is due to inherent
causes only or whether some of the variability is due to assignable causes. This
question is the heart of Statistical Quality Control. If only inherent variation
exists there is little, short of a major change, which can improve the process. If,
on the other hand, assignable causes are present, we should find them and eliminate
them. When assignable causes of variation are discovered and corrected soon after
they occur, we have a means for controlling quality. The purpose of control limits
is to permit us to distinguish assignable causes from inherent causes so that
appropriate action can be taken. The control limits of both Xbar and R are located
so that they will include nearly all of the variation due to chance causes.
Therefore if a plotting on either chart falls outside the control limits, we can be
confident that an assignable cause exists and may be located by investigation.
Occasionally, about three times out of a thousand, a plotting subject to only
inherent variation will fall outside the limits. This occurs so infrequently that
any plotting outside the limits should be adequate reason for an investigation. The
investigation may sometimes be fruitless, but far more often it will reveal the
change necessary to improve the process. In our example we find no points outside
the limits. Therefore we would assume that no assignable causes are present and
that the variation is due to a chance cause system. DETECTING ASSIGNABLE CAUSES In
the above discussion we have pointed out that a plotting outside the control limits
is evidence that an assignable cause exists and action should be taken. There are
other methods for detecting an assignable cause. Although a plotting outside of the
control limits may be an indication of an assignable cause, a plotting inside the
limits doesn't necessarily indicate the presence of only inherent causes. An
assignable cause may be indicated by (1) plottings outside the control limits, (2)
shifts or (3) trends. 1. Plottings Outside the Control Limits If a plotting on
either the Xbar chart or the R chart falls above the upper control limit or below
the lower control limit we say that a plotting this far from the center line is an
indication of an assignable cause. We should seek out and correct the cause of this
excessive variation. 2. Shifts If a series of consecutive plottings fall either
above or below the center line of either chart, we say that it appears that a shift
has occurred. Even though all the plottings are still within the control limits, we
can be quite confident that an assignable cause exists. Again, we should err on the
side of caution and look for an assignable cause in order to avoid the possibility
of producing faulty product. Various rules can be utilized to detect the presence
of shifts, such as 2 successive plottings near either of the control limits, or 3
successive plottings at least one and one-half standard deviations on either side
of the center line, or 7 successive plottings on the same side of the center line,
or 10 of 11 successive plottings on the same side of the center line of the average
chart. 3. Trends If a series of consecutive plottings move steadily either upward
or downward on either chart, we say that this is an indication of a trend. Actually
a trend is a shift which takes place slowly. No specific rules are commonly used in
detecting trends, but a modified interpretation of shifts is adequate. A trend is
adequate indication of an assignable cause. THE MEANING OF "IN CONTROL" A process
which the control charts indicates as not being affected by assignable causes is
said to be "in control." If assignable causes are present, the process is said to
be "out of control." " ." i , " ." " ." erstood that an
assignable cause does not mean that the product is defective nor does a constant
cause system indicate quality conformance. The chart has only chcked for
consistency in the process. There is no consideration at this point for the manner
in which the product conforms or fails to conform to the specifications. COMPARING
THE PROCESS WITH THE SPECIFICATIONS As we have discovered, an objective of
statistical quality control is to provide information, pertinent to quality of
product, which will permit the production of a useable product or a useful service
at the least cost per unit of product or service. This usually leads to the
ultimate improvement of manufactured products. More specifically, in process
control, we want information which will permit us to determine the manner in which
the parts being produced conform to that which is desired -- the specifications. To
accomplish this we sample parts as they are produced and make inferences, or
assumptions, about the universe from which the samples were taken. In production,
the universe is the constant cause system which produces the parts or product. We
are no more interested in the five pieces in the sample than we are in any other
five pieces in the universe. The sample merely provides the data from which we make
inferences about the universe from which they came. We are often tempted to assume
that all of the product conforms to specifications if we see no defective items in
the Xbar-R data sheet. This is not the case because we have taken a sample. We can
be relatively sure that we may not have found the largest or smallest value among
those we have measured. PREDICTING INDIVIDUAL VALUES Statistical theory tells us
that there is relationship between the range and the standard deviation. As we have
pointed out, the range gives a measure of the total spread while the standard
deviation tells something about how the values cluster about the mean. In our
example we calculated the range to describe variation. This was satisfactory as a
long as our samples were of the same size. However, the average range, Rbar, for
our twenty samples was 4.3, yet the 100 values actually varied from -4 to +4. As we
take larger samples, the range increases because we are more likely to get extreme
values as our sample size increases. This is where the standard deviation comes
into play. The value we seek must be one that increases as the variation increases
but is also in the same units as our data. This value is the standard deviation of
x and is represented by the symbol SD(x') read "sigma sub-x prime." The standard
deviation of x is calculated in the following way. SD(x') = standard deviation of x
SD(x') =Rbar/d(2) Rbar = average range d(2) = a factor which depends on the size of
the sample where the sub-x indicates that it is the standard deviation of the
individual x values, and the prime identifies it is the standard deviation of the
universe from which the samples are taken. An estimate of SD(x') is Rbar/d(2)
provided the process is consistent, or in control. There are other methods of
calculating the standard deviation; however, the above formula is used almost
exclusively in statistical quality control work. The control chart data was
obtained from a sample of five. From the table "Factors for Control Limits" found
on the reverse side of the ASQC VARIABLES CONTROL CHART, we find that d(2) = 2.326.
Therefore, SD(x') =Rbar/d(2) = 4.30/2.326 = 1.85 Let's calculate the value of
Xdouble bar 1SD(x'). Xdouble bar + 1SD(x') = +.066 + 1.85 = +2.51 Xdouble bar -
1SD(x') = +0.66 + 1.85 = -1.19 From the table "AREAS UNDER THE NORMAL CURVE" we
find that .3413, or 34.13% of the area of the normal curve is between 0 and 1.00
standard deviation units. Therefore, we can predict that approximately 34% of the
measurements will be between +0.66 and +2.51. We can also predict that
approximately 34% of the measurements will be between -1.19 and +0.66. It
would follow then that approximately 68% will fall between -1.19 and +2.51. This
is shown in the following graph. (INSERT NORMAL CURVE with 1 sigma either side of
the average) Let's see what proportion will fall between 2 standard deviations.
Xdouble bar + 2SD(x') = 0.66 + 2(1.85) = + 4.36 Xdouble bar - 2SD(x') = +0.66 -
2(1.85) = -3.04

You might also like