Sunteți pe pagina 1din 31

Understanding

2005 LUDECA, INC.

Standard Deviation

Unbiased Standard Deviation

( x i x)
SD
n

1
i1

2005 LUDECA, INC.

Are you intimidated yet?


The whole point of this presentation is to unintimidate you
regarding this excellent feature of the Rotalign Ultra.

www.ludeca.com

2005 LUDECA, INC.

Now you see the Standard Deviation Result in the same screen. You can enable or
Here you see alignment results in the Measure screen, after readings
disable the appearance of this information at will. However, we recommend that you
have
been
information.
[Please
click]
do
enable
it. taken,
This is awithout
criticallyStandard
importantDeviation
and very powerful
feature
that only
your
Ultra has, and it can make the difference in achieving certain critical alignments.

www.ludeca.com

2005 LUDECA, INC.

To turn on the SD display, go to the Global Menu and select Units.


[Please click.]

Then select Show on Measurement screen in the Showing SD value


window.

www.ludeca.com

2005 LUDECA, INC.

Even if you have SD turned off in the main Measure Screen, you can
still see the SD value in the Measurement Table.
Just scroll to the right a little.

www.ludeca.com

Understanding Standard Deviation

2005 LUDECA, INC.

So then, what is Standard Deviation?


Standard Deviation is the mean of the means. It
describes how closely a group of data points are
clustered around the average of those data points.
It is a measure of measurement quality. The smaller
the SD, the better the quality of the data you
collected.
www.ludeca.com

Understanding Standard Deviation

2005 LUDECA, INC.

On a very critical alignment, or where you have a lot


of vibration, or couldnt turn the shafts, just getting
the results may not be good enough. Of course, a
repeatability check is essential, but even if you
pumped up the averaging of the readings in
Multipoint, are you really sure the data is good??
With Standard Deviation display, the need for a
repeatability check disappears, which can save you
time under difficult conditions.
www.ludeca.com

Understanding Standard Deviation


The alignment results you get are the average of all the data
points collected. Why is this average of the data, by itself, not
good enough, even if it is repeatable? [Please click]

2005 LUDECA, INC.

Suppose youre planning a party for a group of kids. All you are told is:
the average age of the group is 10.

Look at the ages: scattered all the way from 2 to 20. Ask yourself: How
successful is my party going to be if I plan it around activities suitable
for age 10?

www.ludeca.com

Understanding Standard Deviation

2005 LUDECA, INC.

Obviously, your party will be a bomb. Lets look at the distribution of ages again: [Please click]

If you had been told the


standard deviation of the
group, youd know you were
in trouble. [Please click]

www.ludeca.com

Understanding Standard Deviation

2005 LUDECA, INC.

Now suppose you had this group instead:

All the kids are much closer


in age to the average, so your
party will be a hit! Look at the
Standard Deviation: [Please click]
Much better! As you can see,
if all youre told is the average
of the data, it doesnt tell you anything about the quality of the data.

www.ludeca.com

2005 LUDECA, INC.

Understanding Standard Deviation


In perfect alignment, the laser
beam describes a dot on the sensor that never moves as you rotate
the shafts. But when misalignment
exists, the beam will move in an
ever changing arc. Each dot represents a reading taken somewhere
on this arc. The dots are not all
positioned perfectly upon the arc.
Some fall above or below it. If we
average out all their positions, we
get a curved path that represents
the average trajectory of all the
dots. Eventually, over a full rotation of the shafts, this averaged
beam track describes an ellipse.

www.ludeca.com

2005 LUDECA, INC.

Understanding Standard Deviation


However, just looking at this
average is not good enough,
because, like the ages of the
kids, it doesnt really tell us
much about how good the data
is, or how far above or below
the average trajectory the
points fall. Clearly, if we are
deriving the trajectory of the
arc from the average, and some
points are widely scattered,
that average will be greatly
affected, and it would be good
to know that.

www.ludeca.com

Understanding Standard Deviation

2005 LUDECA, INC.

If we take the
ellipse, cut it,
and lay it out
flat, we get the
deviation diagram shown here.
This is called
Broken Ellipse
view in the Ultra.
Notice that one
point seems to
be way out of
the pattern or
average track
of the others.
Clearly, this particular point represents an aberration of some sort.

www.ludeca.com

Understanding Standard Deviation

2005 LUDECA, INC.

Taking the
aberrant point
into consideration will affect
the calculated
results harmfully. It would
skew the average. It should
be deleted from
consideration.
How would you
ever know this
is happening?
A high Standard Deviation value would immediately alert you.

www.ludeca.com

2005 LUDECA, INC.

Using Standard Deviation


This is the result of a set of
Multipoint readings taken on
a large gas turbine. While taking the readings, the result
values were stable and looked
great, but all of a sudden the
numbers changed and the SD
value jumped up. Since rotating
the shafts was very difficult,
it was not convenient to start
over again. Since the numbers
had been good up to that point,
it was decided to continue since there was nothing to indicate that the laser or
receiver had been bumped, or had hit anything. After 33 points were taken the
rotation of the shafts was finally completed and readings were stopped. Clearly,
with an SD of 53, something had gone seriously wrong, yet the prospect taking
another set of readings was to be dreaded. To see what went wrong with the
readings, we went to the Menu and...

www.ludeca.com

2005 LUDECA, INC.

Using Standard Deviation

... selected Edit points.

[Please click]

www.ludeca.com

2005 LUDECA, INC.

Using Standard Deviation

Broken Ellipse view is selected. Note the deviation diagram in this


screen. Clearly, at least one of those 33 readings was very bad. Next,
instead of scrolling to it, just press Menu again and...

www.ludeca.com

2005 LUDECA, INC.

Using Standard Deviation

...select Max point. This will automatically select the point with the
highest deviation in the group when you press the
key. [Please click]

www.ludeca.com

2005 LUDECA, INC.

Using Standard Deviation

Simply press the


key to disable the point. Next, youll see what
happened when the worst point was disabled.

www.ludeca.com

2005 LUDECA, INC.

Using Standard Deviation

The SD dropped all the way down to 2.19! You can continue disabling
points the same way until you are satisfied that your SD value is low
enough that you can rely on the quality of the remaining data.

www.ludeca.com

2005 LUDECA, INC.

Using Standard Deviation


One important thing to
look at in deciding whether or not to disable a
point is how far away
from the trajectory or
average it lies. Look at
the Delta value. If this
number is fairly high,
chances are great
youll improve the SD.
Also make sure that you
dont disable too many
points. It is important
that the number of
remaining active points
be sufficient to still
render your data meaningful. (By the way, the aberrant worst point was caused by
taking that point just as the laser beam barely clipped the chain fall cable that was
being used to turn the shafts. When disabling the nine points with the highest
deviations we worked with the best data and lowered the SD to just 0.8 mils!)

.
www.ludeca.com

2005 LUDECA, INC.

Here we see the


laser and
receiver at
about the 12
oclock position.
Note that the
beam is shot
through the
coupling bolt
hole.
After the shafts
were turned, the
laser beam just
barely clipped
the chainfall
cable at that one
point, located at
about 10 oclock.
This can be seen
in the Edit
Points screen
that we saw
earlier.
[Please click]

www.ludeca.com

2005 LUDECA, INC.

So... When should I disable a


point??

The red point is clearly not on the standard track of the others.
It is an aberrant point and should be disabled.
The Standard Deviation will improve significantly.

www.ludeca.com

2005 LUDECA, INC.

So... When should I disable a


point??
The red point has the
largest absolute deviation, but, ask yourself, is it really furthest away from the
average track of
the others? Would a
straight line average
track really represent
what these points are
collectively trying to
tell us? No! Here the
sinusoidal track appears to be a more accurate representative of reality.
Disabling this point will not significantly change the Standard Deviation!
This illustrates that the average itself, and the absolute deviation from the
average mean little, whereas the Standard Deviation is very significant.
You should not disable this point.

www.ludeca.com

2005 LUDECA, INC.

So... When should I disable a


point??
Again, the red point
is furthest away from
the average track of
the others, but is it
really so bad compared to all the others?
Will disabling it really
affect what all the
other points are trying to tell us? Again,
you will find that disabling this point will
not reduce the Standard Deviation much. Since the data is quite rough, more
points are what is needed to ensure good reliable data. Collectively, the
Standard Deviation of all of those points may not be that bad. Again, choose
not to disable the max point in this case.

www.ludeca.com

2005 LUDECA, INC.

Standard Deviation & Repeatability

Note the poor repeatability of these readings, and the high


standard deviation of the first set.

www.ludeca.com

2005 LUDECA, INC.

Standard Deviation & Repeatability

Note that all points taken


are active in the first set of
readings that has the high
Standard Deviation.

www.ludeca.com

2005 LUDECA, INC.

Standard Deviation & Repeatability

Now look at what happens


when we disable the worst
points and bring down the
Standard Deviation: The
repeatability is now excellent!

www.ludeca.com

2005 LUDECA, INC.

Standard Deviation & Repeatability

This proves that if your Standard Deviation is low, your results


will be accurate and you do not have to take another set of
readings to establish repeatability. Save time!

www.ludeca.com

Standard Deviation Tolerances

For normal applications:

0 SD 1.0 mils

For critical alignments:


2005 LUDECA, INC.

0 SD 0.5 mils

www.ludeca.com

2005 LUDECA, INC.

Standard Deviation Bell Curve

This is the SD bell curve. What this shows is that in a normal distribution
of measured values, 68.2% of your data will always fall within one SD of
the mean, and 95.4% of your data will fall within 2 SDs, and 99.6% of the
data will fall within 3 SDs. The important thing to remember is that the
actual value of the SD should be kept low.

www.ludeca.com

S-ar putea să vă placă și