Saturation Curve Analysis and

Quality Control by David Kirk

INTRODUCTION
Users specify the range of indentation ability of the shot streams
to be used on their components. They are able to do this by virtue
of the so-called “Saturation Intensity” which is a quantitative
measure of shot stream indentation ability. The range is normally
specified as upper and lower limits for the saturation intensity,
which has to be obtained from a saturation curve. At various
stages in shot peening we need to confirm that the specified
indentation ability is being employed. These stages include the
initial set-up of a job and subsequent verification intervals. The
primary quality control application of saturation curve analysis
is, therefore, the determination of saturation intensity values.
A secondary application is to obtain an indication of the Almen
strip indentation rate.

SATURATION CURVE SHAPE

Each shot particle that indents the surface of an Almen strip
causes a minute plastic expansion of that surface. This expansion
induces a corresponding tiny increment of convex curvature into
the strip. Because a peened strip has received a very large number Fig.1 Effects of shot velocity and flow rate
of indenting particles we get a measurable curvature – expressed on position of ‘saturation point’.
as the deviation from original flatness and termed “Almen Arc curve there is only one such point – shown as dots in fig.1. It
Height”. On initial exposure to a constant shot stream each shot should be emphasized that the saturation point is not a data
particle can impose a similar increment of curvature. As a point, it is a derived point. The saturation intensity is a defined
consequence the Almen arc height initially increases almost high-curvature point of the saturation curve. There are alternative
linearly with peening time. With further peening, the strip surface ‘characteristic points’. Mathematically-minded readers will note
progressively work hardens so that the tiny increment of curvature that the curve's curvature at the ‘saturation point’ is close to the
attributable to each indenting particle is reduced. The rate of ‘point of maximum curvature’. If we know the mathematical equa-
Almen height increase must therefore slow down. Eventually the tion for the curve we can derive the point of maximum curvature
incremental contributions become negligible. The slowing down by solving a relationship that includes the first and second deriva-
and subsequent leveling-out are the reasons for the characteristic tives of the curve’s equation.
shape of Almen saturation curves. There are various specifications that detail the requirements
Shot streams with different indenting ability will give for saturation curve measurements. All of these specify that several
different ‘saturation curves’. With increase in shot velocity (and Almen strips must be exposed for different times to the same shot
therefore of indenting ability) there is a corresponding increase in stream. The measured arc heights are then plotted against peening
curve height, see fig.1. We should also note that the greater the time. A curve must then be drawn so that the saturation intensity
shot flow the quicker will be the increase in arc height. That can be estimated. There are two alternatives: manual curve fitting
means that we can have different saturation curves without any and computer-based curve fitting. With the universal availability
difference in indentation ability. of computers and appropriate curve-fitting procedures the former
Consider next the problem: “How can we assign to each satu- technique should be ‘consigned to history’.
ration curve a quantity that uniquely defines the indenting ability
expressed by that curve?” To solve that problem we need to find a SATURATION CURVE PREDICTION
particular point of the curve that defines the curve. The standard One advantage of computerized curve-fitting is that the curve’s
solution is the so-called "ten percent rule”. This solution gives us: equation has parameters that are directly related to saturation
“The (first) point on the curve for which doubling the peening intensity and saturation time. Popular equations used for curve-
time increases the arc height by 10%”. For every saturation fitting are ‘two-parameter exponential’ and ‘two-parameter
saturation growth’. These are:
h = a(1 – exp(-b*t)) (1)
Dr. David Kirk, our “Shot Peening Academic”, and h = a*t/(b + t) (2)
is a regular contributor to The Shot Peener. where h is arc height, t is peening time, a and b are the two
Since his retirement, Dr. Kirk has been an parameters.
Honorary Research Fellow at Coventry For both equations the saturation intensity is a fixed proportion
University, U.K. and is now a member of of parameter a, [9a/10 for equation (1) and 9a/11 for equation (2)].
their Faculty of Engineering and Computing. Similarly the saturation time is a fixed proportion of parameter b,
We greatly appreciate his contribution to our
[2·303/b for equation (1) and 4·5*b for equation (2)]. Fig.2
publication.
Continued on page 26

Summer 2006 24 The Shot Peener

SATURATION CURVE ANALYSIS AND QUALITY CONTROL Case Study: Attempt to produce an
Continued from page 24
Almen Arc height of 0.015'' using
S110 steel shot.
The author’s records are stored as Excel spreadsheets with
separate sheets for each type and size of shot. For each shot
type there are separate columns for air pressure, shot feed rate
(with actual MagnaValve settings), gun-to-component distance,
gun type (suction fed or direct fed), angle of impact, Almen
strip type, test date, arc heights and times, saturation intensity,
saturation time and fitted curve parameters. Complementary
sheets detail the history of each shot charge in terms of origin,
purchase date, sieve details, image analysis, etc. With only
about a thousand saturation curves on record it was a simple
task to use Excel’s Data/Sort facility to highlight the several
previous combinations of settings that yielded approximately
0.015” when using S110 steel shot. The recorded saturation
curve parameter values were then fed into a Curve Solver
computer program to produce a “reference curve”. The
combination of machine settings that gave that “reference
curve” were then used to produce a new, current, saturation
curve. This new curve was then plotted on the same graph as
the reference curve. Fig.3 shows the outcome.
Fig.2 Effect of arc height and peening time on two-parameter
exponential saturation curves. Current Curve
a b Sat.Int. T
illustrates these relationships for equation (1). The similarity with 15.31 0.43 13.78 5.32
fig.1 is not coincidental!
The machine control settings that lead to every saturation Reference Curve
curve produced by a particular peening shop should have been a b Sat.Int. T
documented. Settings for a new job can therefore be based either 16.49 0.32 14.84 7.20
directly on past records or on the superintendent’s wealth of expe-
rience (or both). Armed with a knowledge of the equation parame-
ters we can plot an expected saturation curve immediately. The
case study shown in the next column illustrates the approach used
by the author for his laboratory peening facility.
The primary factors that govern saturation curves, for a given
shot charge, are shot velocity and shot stream flux. In this context,
‘flux’ is the number of shot particles crossing each unit area of the
shot stream’s cross-section. Shot velocity is controlled by varying
either air pressure or wheel speed. Shot stream flux is varied by
means of some type of feed valve – such as a Magnavalve. There
is, however, an inter-dependence of shot velocity and shot flux.
That means that we cannot vary velocity and flux independently.
There are several factors that contribute to the inter-dependence.
The major factor is the efficiency of energy conversion. For an
air-blast machine the compressed air is providing kinetic energy, Fig.3. Comparison of reference and new saturation curves
some of which is translated into kinetic energy of the shot parti- using Curve Solver 2EXP v3.00C.
cles. The greater the shot flux, the lower is the air stream’s For the situation represented in fig.3 there is a clear discrepancy
efficiency in accelerating the shot particles. Complex physics between the two curves. The machine settings have produced a
are involved! lower saturation intensity and shorter saturation time than was
expected. By cross-checking with the complementary data it
DATA POINT SELECTION was found that the S110 shot charge had been in use for so
Data point selection for a saturation curve is very important, but is long that its average diameter was substantially lower than
rarely mentioned in specifications. The primary objective is to when the reference curve was produced. That led to a lower
determine the characteristic ‘saturation point’ to within reasonable intensity potential and a faster coverage rate. The problem was
confidence limits. It follows that the range of selected data points subsequently solved by clearing out the shot charge and
should straddle the expected saturation point without being too replacing it with new shot.
close together. This means that we should have some points at
lower times than the saturation time and some at longer times. The
most useful sequence of peening ‘times’ is generally found to be
based on a ‘geometric progression’. For example, the sequence of
numbers 1, 2, 4, 8, 16 and 32 is a geometric progression where

Continued on page 28
Summer 2006 26 The Shot Peener
SATURATION CURVE ANALYSIS AND QUALITY CONTROL
Continued from page 26
each number is double the preceding number. Such a sequence
normally allows an efficient utilization of the limited number of
data points.
Fig.4 illustrates a real situation where the choice of data
points was not efficient. This situation occurred because a high
shot stream flux was imposing a very high coverage in a single
pass over the Almen strips. Computer-based curve-fitting yielded
the indicated T and 2T arc heights. Typical error bars are shown
for the four data points. It takes little imagination to appreciate
that, given the error limits shown, the ‘real’ T value is anywhere
between 0·1 and 1·0 strokes. We cannot, however, select fractions
of a peening stroke! Stroke speeds can often be increased to yield Fig.4 Almen saturation curve with data points beyond the
the equivalent of stroke fractions. With fig.4, half and quarter saturation point.
stroke fractions would be far more effective than the three and deviations from the mean. Confidence limits will only be main-
four stroke points. It would be simplistic to argue that the shot tained for actual measurements if there is no change in mean arc
feed rate should be reduced so that the saturation time becomes height. Again that means that we must control both shot velocity
much longer. Shot feed rate reduction would mean that production and shot stream flux.
rates would suffer badly. Multi-block T-testing is more complicated than single-block
T-testing. Some components require an array of several Almen
COVERAGE blocks – in extreme cases more than twenty – for the set-up
Saturation times are a useful guide to the rate at which compo- saturation intensity determination. Intermittent confirmation test-
nents will receive specified coverage levels. For any given ing may then require single strips to be tested at the same time at
machine set-up, the coverage rate of the Almen strips is inversely all of the locations. That ‘same time’ cannot correspond to the
proportional to the saturation time. Hence, the shorter the satura- saturation point ‘time’ for all of the test blocks. Each block will
tion time the higher will be the coverage rate. The coverage rate have yielded a different saturation ‘time’. One reasonable way of
for a component will not be the same as that for Almen strips. handling the situation is to take the mean of the saturation times
That is because there is normally a difference between the indenta- derived for all of the locations involved and require that the
tion resistance (hardness) of the component and that of the Almen nearest integral ‘time’ be used - with compensating adjustments
strip. The coverage rate for a component is therefore inversely to the required arc height limits.
proportional to both saturation time and component hardness. If Two-strip T-testing is required by some users. This involves
the component is softer than the Almen strips then it will receive a tests being carried out at two different peening times, T and 2T,
higher coverage rate than will the strips. Conversely, if the compo- with a requirement that the arc height at 2T will be less than 10%
nent is harder than the Almen strip then the component’s coverage greater than that at T. The test is more rigorous and more compli-
rate will be lower than that for the strips. cated to analyze than single-strip T-test confirmation.

T-TESTING DISCUSSION
T-testing is an important feature of quality control. The objective Every full Almen saturation curve is a confirmation test of a shot
is to check periodically that the shot stream's indentation ability stream’s required indentation ability and is therefore an essential
is being maintained after the original set-up curve has been part of quality control. The curve yields an internationally-accepted
produced. parameter, the so-called “saturation intensity”, whose derivation is
Single-strip T-testing is a straightforward test with the clearly- classically simple – using either manual or computerized curve-
defined requirement that the measured arc height for a strip peened fitting procedures. Saturation intensity is a high-curvature point of
for a designated time, T, has to be between stated upper and lower the saturation curve – it is not a data point. Continued on page 30
limits. The test may require either one or more strips to be tested
for one or more Almen block locations. If the test is to be effective
then control has to be exercised over both shot stream flux and
shot velocity. In practical terms we have to control both shot feed
Almen Saturation Curve
rate and air pressure/wheel speed. Solver Program
The measured arc height on a single strip will rarely be
precisely the same as that predicted from the full saturation curve FREE from The Shot Peener
test. That is because all measurements have unavoidable variability –
which can be expressed as a standard deviation. Every Almen arc
height measurement is a ‘statistic’ and collections of statistics are
Get the program developed
best treated using reliable, appropriate, statistical techniques. The by Dr. David Kirk
commonest statistical parameter is the ‘average’ of a collection of Request the program at:
values. A ‘normal distribution’ of values has two parameters: the http://www.shotpeener.com/learning/solver.htm
average or mean, µ, and the standard deviation, σ, of the values.
‘Confidence limits’ are defined as the probability that a measure-
ment will lie within those limits. Hence ‘95% confidence limits’
Over 400 copies have
would be plus and minus two standard deviations from the mean. been downloaded!
’99·7% confidence limits’ are plus and minus three standard

Summer 2006 28 The Shot Peener

SATURATION CURVE ANALYSIS AND QUALITY CONTROL
Continued from page 28
Saturation time can be used as an indication of the shot stream’s The Original CLELAND
coverage rate. With computerized curve fitting procedures the
derived saturation point is determined objectively and it is possible SHOT CLEANING SPIRAL SEPARATOR
to quantitatively relate actual curves to those that would be antici-
pated.
The objective with T-testing is to confirm that the required The Cleland Spiral Separator
saturation intensity is being maintained during a production run. is the most successful method
It cannot, however, guarantee that this intensity is being main-
of separating round product
tained. That is because different saturation curves can intersect at
the specified saturation point. Consider, for example, the situation from cracked, broken, or non-
illustrated by fig.5. A set-up saturation curve is shown, together round materials. The Cleland
with the corresponding saturation point, Ts. A second saturation Spiral Separator is available
curve is shown which could easily have arisen during a production with a self-cleaning hopper to
Spiral
run if the shot flow rate had been substantially reduced – hence Separators ensure total emptying of the
inducing increased shot velocity (if air-blast is being employed).
The second curve has the same Almen arc height at Ts as the set-
top hopper bin.
up curve, but has a different saturation intensity, Tc. Hence, a
confirmation T-test would not reveal that the saturation intensity
had in fact substantially increased. “Cleland Spirals Work Around the World”
Two-strip T-testing is difficult to quantify reliably. It cannot
be either as effective or as reliable as a full saturation curve. The
use of full saturation curves should, therefore, be the preferred Phone/Fax: (763) 571-4606
practice, especially for critical applications.
Cleland Manufacturing Company
2125 Argonne Drive
Minneapolis, Minnesota 55421 USA

SAM YOUNG
ABRASIVES MEDIA
Carbon & Stainless Steel Cut Wire Shot
Carbon & Stainless Steel Conditioned Cut Wire Shot
Fig.5 Identical arc heights obtained at Ts with Aluminum Cut Wire Shot
different saturation intensities. Zinc Cut Wire Shot
CONCLUSIONS Zinc Conditioned Cut Wire Shot
1. Saturation intensity is a reliable, primary, quality control Zinc Cast Ball Shot
parameter. It is a high-curvature point of the saturation curve –
not a data point.
2. Computerized curve-fitting has substantial advantages relative SAM YOUNG INDUSTRIAL CO. LTD.
to manual curve-fitting.
3. Single-strip T-testing for confirmation testing is useful provided 186 HYO SUNG DONG
that effective control is exercised over both shot flux and shot
velocity. KYE YANG KU
4. Saturation curves can provide a useful indication of coverage
if an appropriate allowance is made for the relative indentation
INCHON, KOREA
resistance of Almen strip and component material. TEL: +82-32-547-0357
FAX: +82-32-548-9350

E-MAIL: samyoungshot@hanmail.net

Summer 2006 30 The Shot Peener