Picking up where we left off,
we seek to establish a natural pattern.
Natural patterns are indicative of a stable system of common cause variations.
This is one where first,
2/3 of the points are near the center value
or a few of the points are on or near the center value.
Points appear to float back and forth across the center line.
The points are balanced in roughly equal numbers on both sides of the center line.
There are no points beyond the control limits.
And finally there are no patterns or trends on the chart.
Here is an example of a natural pattern.
Some typical signs that are process is out of control include
values outside the control limits, trends or cycles.
There are many other conditions we must look out for besides these three
common behaviors.The center line is where the average or target value will reside.
If the process is under statistical control then we
can estimate process parameters: mean,
standard deviation and process capability.
Per the Central Limit Theorem,
as we increase the sample size this causes
the control limits to be drawn closer together,
since it decreases the variance.
Control limits are usually placed at plus or minus three sigma.
This captures 99.74% of the behavior of the sample statistic.
This is our indicator.
The process is in control and our process is consistently performing within these
limits.Programs such as minitab screen for the full scope of behaviors we must look for.
Each role has approximately the same probability of occurring.
Other SPC programs offer comparable checks.
One thing to be careful with regarding
these rules is that you may not want to react too quickly.
Several factors need to be considered.
First are consequences high if we do not
react to a potential unstable system using the 8 rules?
Two, can we potentially make a process unstable if we react too
quickly to the control chart when in reality that process is not out of control?
In other words don't fix it if it's not broken.
And finally some suggest using only 1-3 of these rules for stability of the process.
Using only a portion of these rules will minimize
the potential instability caused by tampering with a stable system.
However you could be potentially reacting to false alarms.
Obviously, if we see one or more points above the three sigma limit we have an issue.
If we see nine points on the same side of the center line,
this indicates that the central tendency of the process is drifting.
We can confirm this by looking at the R chart where applicable.
If we see an increase then the process is unstable.
If we see six or more points increasing or decreasing,
this constitutes a trend.
This would also indicate a problem.
14 points alternating up and
down is indicative of a process capability that is diverging.
These 14 points are no doubt the beginning of the end regarding process stability.
If we see two of three points in a row
outside the plus or minus two sigma level on the same side,
we may expect a large fluctuation or the onset of a divergent process.
4/5 Points outside the one sigma limit
on the same side would indicate bunching or clustering.
Such behavior is most common in X bar,R charts or P charts.
15 points within one sigma of the center line are indications of stratification.
Look at the R chart to best see this behavior.
When we have eight points in a row,
more than one segment from the center line on either side, this indicates cycling.
So where can process variation come from?
It may be simply long term.
It can also be attributed to the lot,
stream, time or the piece.
It can even be within the piece itself especially for complex components.
There may also be an error in
the measurement or shortcomings regarding process capability.