CHAPTER 6

DIMENSIONAL ACCURACY ANALYSIS

The previous chapters discussed how six different input parameters affect the surface
roughness and hardness of ABS composite implants. Mathematical models have been developed
to achieve the desired surface properties. Although the vapor smoothing process improves the
surface finish and hardness of ABS replicas, it can cause dimensional changes. This happens
because partial melting and the repositioning of layers can affect dimensional stability in various
areas. Since these replicas are intended for industrial use, it's crucial to control these dimensional
changes. Dimensional accuracy is important for manufactured components as it shows how
effective the manufacturing and finishing processes are. For composite implants, accuracy is
especially important because even small variations can cause pain, discomfort, and joint
dislocation after surgery (Sorrentino et al., 2010).

6.1 MEASUREMENT OF DIMENSIONAL ACCURACY

In this chapter, we aim to analyze the key process parameters of FDM and vapor
smoothing (VS) that cause dimensional changes in composite implant replicas. We focused
on three critical dimensions: head diameter, neck thickness, and stem thickness (Figure
6.1). We used a Coordinate Measuring Machine (CMM) to measure these dimensions at
three points before and after the vapor smoothing process. To generate experimental data
for these critical dimensions, we used the Taguchi L18 Design of Experiments (DOE)
method. We then analyzed the Signal-to-Noise (SN) ratio plots to see how different
parameters of the combined FDM-VS process affect the dimensions. Response tables of SN
ratios helped us create equations to predict the results at optimal levels. An ANOVA table
was used to determine the relative importance of each parameter with 95% confidence,
showing the percentage contribution of each one. We validated the results using residual
plots for each critical dimension. In present study, the percentage deviation in dimensions
has been considered as a response so as to minimize the error during calculation of SN
ratios. It is calculated as:
 original ( CAD ) dimension−actual dimension
Percentage Deviation=
original ( CAD ) dimension

12.0524
Location 1 (Head Diameter)

1.5 7.7458
Location 2 (Neck Thickness)

Location 3 (Stem Thickness)

6.620

All dimensions in mm
74

Figure 6.1 Location and dimensions of three critical part features

We obtained the original CAD dimensions of each part feature from "Solidworks 2014"
and compared them with the actual dimensions. We calculated the percentage deviation
twice: before and after the vapor smoothing process.

To handle uncontrolled noise factors, we used the Signal to Noise (SN) ratio to select a
robust experimental design. The SN ratio measures how sensitive the response is to
controlled conditions compared to uncontrolled external noise factors. In this case, we
calculated the SN ratio using the "smaller is better" quality characteristic because our goal
is to reduce dimensional variability (percentage deviation). The formula for calculating the
SN ratio was applied accordingly.
 S 1 2
=10 (∑ y ) ….. (6.1)
N n

Where y is output response and n is number of observations.

The IT Grades for these three critical dimensions were calculated for all eighteen experiments
according to the UNI EN 20286-1 standards. The standard tolerance unit \( n \) for the nominal
dimension TJN was calculated as follows:

𝑛 = 1000 (TJN – TJM) / 𝑖

Where TJM is a measured dimension and 𝑖 is standard tolerance factor The

standard tolerance factor 𝑖 was calculated using formula:

𝑖 = (0.45 x T1/3) + (0.001 x T)

Where T is the geometric mean of the nominal dimension in mm

Therefore, the IT grade for each experiment can be determined from the IT grade chart
(Appendix II) based on the standard tolerance unit. A higher IT grade value indicates a larger
tolerance according to industrial standards. For the measurements of the three critical features of
composite replicas, the IT grades are listed in the last column of the observation tables, ranging
from IT grade 4 to 13.
In this experiment, the percentage deviation was calculated both before and after smoothing.
However, for statistical analysis, we focused on the percentage deviation after smoothing to
understand the impact of the vapor smoothing (VS) process on dimensions. The percentage
deviation before smoothing (%ΔDb) was calculated as follows
original CAD dimension−actual dimension before smoothing
% D b= ×100
originalCAD dimension
Similarly, the percentage deviation after smoothing (%ΔDa) has been calculated as:

original CAD dimension−actual dimension before smoothing

% ∆ D= ×100
original CAD dimension
6.1.1 DIAMETER MEASUREMENTS AND ANALYSIS

The table 6.1 lists nine objects along with their initial and final diameters, as well as the
percentage change in diameter.

S.No. Initial Diameter Final Diameter Percentage Change

1 9.31 9.34 -0.32223
2 9.2 9.24 -0.43478
3 8.94 9 -0.67114
4 9.44 9.14 3.177966
5 9 8.58 4.666667
6 9.51 9.34 1.787592
7 9.7 9.42 2.886598
8 9.54 9.62 -0.83857
9 9.51 9.3 2.208202

Table 6.1 Measurements of diameter before and after smoothing

For each object, the percentage change column indicates how much the diameter changed
between the initial and final measurements. A negative percentage change means the diameter
decreased, while a positive percentage change means it increased. For example, object 1's
diameter decreased by 0.32% from the initial measurement to the final measurement. It's
important to note that the table doesn't specify the units of measurement for the diameters, nor
does it explain what caused the changes in diameter.
 Figure 6.2 Graph for measurement of Temperature, Time, and RPM

The image is showing data from an experiment involving signal-to-noise (SN) ratios. SN ratio
compares the strength of a desired signal to background noise, with a higher ratio indicating a
stronger signal and lower noise level. In this experiment, the phrase "Smaller is better" next to
"Signal-to-noise:" means that a lower SN ratio is preferable for the outcome. The figure 6.2
illustrates how temperature and time affect the SN ratio, with the temperature, time, and RPM
(revolutions per minute), and the mean of SN ratios. For all three RPM levels (1880, 2490, and
2670), the graph suggests that the SN ratio generally decreases as temperature increases. The
impact of time on the SN ratio seems more complex and depends on the temperature.

Sr. no Temperature Time RPM

1 -6.8337 -7.7564 2.2082
2 -4.7893 -2.1370 -6.1212
3 0.4269 -1.3028 -7.2831
Rank 2 3 1

Table 6.2 Response for Signal to Noise Ratios

The table 6.2 titled "Response Table for Signal to Noise Ratios" illustrates the relationship
between temperature levels and the time taken for a signal to reach a certain noise level. The
"Temperature" column lists the specific temperature values for each level. The "Time" column
indicates the time required for the signal to reach the noise level at the given temperature, with
lower times being better, as denoted by the caption "Smaller is better" at the top of the table. The
"RPM" column shows the revolutions per minute (RPM) at which the experiment was
conducted. The "Rank" column ranks the levels from 1 (best) to 3 (worst) based on the time
taken to reach the noise level. For example, Sr. no. 2 has the lowest time (-7.7564) to reach the
noise level at a temperature of -4.7893 degrees Celsius and an RPM of -6.1212, making it the
best (Rank: 2) among the three levels. Overall, the table suggests that lower temperatures
generally result in faster times to reach the noise level in this experiment.

Adj
Source DF Seq SS Adj SS MS F P
Temperatur 2 84.10 84.10 42.052 5.42 0.156
e
Time 2 73.92 73.92 36.961 4.77 0.173
RPM 2 160.81 160.81 80.407 10.37 0.088
Residual 2 15.51 15.51 7.755
Error
Total 8 334.35
Table 6.3 Analysis of Variance for SN ratios

ANOVA (Analysis of Variance) is a statistical method used to compare means

between groups in table 6.3. The "Source" column lists the factors being studied,
which in this case are temperature, time, and residual error (unexplained variability).
"DF" (Degrees of Freedom) indicates the number of independent pieces of
information in the data set. "Seq SS" (Sequential Sums of Squares) measures the
variability between groups. "Adj SS" (Adjusted Sums of Squares) is an adjusted
version of Seq SS that considers the number of groups. "Adj MS" (Adjusted Mean
Squares) is calculated by dividing Adj SS by its corresponding DF. The "F" (F-
statistic) compares the variances between groups with the variance within groups. The
"P" (p-value) indicates the probability of observing a test statistic as extreme as the
one calculated, assuming the null hypothesis is true. A lower p-value suggests a more
significant effect.
 Observations from the table show that temperature and time have p-values
greater than 0.05, so we fail to reject the null hypothesis for these factors. This means
there is no strong evidence that temperature or time significantly affect SN ratios.
However, the p-value for RPM is less than 0.05, indicating a significant effect on SN
ratios. It is important to note that this table provides a partial view of the ANOVA
analysis and does not include the total sum of squares or total degrees of freedom,
which are also essential components of ANOVA.

6.1.2 THICKNESS MEASUREMENTS AND ANALYSIS

S.No. Initial Thickness Final Thickness Percentage Change

1 3.6 3.52 2.222222
2 3.64 3.62 0.549451
3 3.84 3.6 6.25
4 3.8 3.64 4.210526
5 3.51 3.82 -8.83191
6 4 3.94 1.5
7 3.8 3.56 6.315789
8 3.8 2.8 26.31579
9 3.91 3.7 5.370844

Table 6.4 Measurements of thickness before and after smoothing

The table 6.4 provides information about the thickness of an object before and after
measurement, along with the percentage change in thickness. "S.No." likely represents the serial
number of the measurement. "Initial Thickness" indicates the thickness of the object before
measurement, while "Final Thickness" shows the thickness of the object after measurement.
"Percentage Change" represents the difference between the initial and final thickness, expressed
as a percentage. Negative values indicate a decrease in thickness, whereas positive values
indicate an increase. For instance, in row 1, the object with serial number 1 had an initial
thickness of 3.60, a final thickness of 3.52, and a percentage change of -2.22%, indicating a
decrease in thickness.
 Figure 6.3: Main Effects Plot for SN Ratios
The figure 6.3 titled "Main Effects Plot for SN Ratios" displays data from an experiment on
signal-to-noise (SN) ratios, where a higher ratio indicates a stronger signal with lower
background noise. The caption "Smaller is better" next to "Signal-to-noise:" signifies that a
lower SN ratio is preferred for the outcome. The graph illustrates the effects of temperature,
time, and RPM (revolutions per minute) on the SN ratio. The x-axis represents temperature,
time, and RPM, while the y-axis shows the mean SN ratios. There are three data series, each
corresponding to a different RPM level (1880, 2490, and 2670). The graph generally indicates
that the SN ratio decreases as temperature increases across all RPM levels. However, the impact
of time on the SN ratio appears more complex and varies depending on the temperature.

Analysis of Variance for SN ratios

Seq
Source DF SS Adj SS Adj MS F P
Temperatu 2 58.28 58.28 29.139 3.76 0.210
re
Time 2 62.76 62.76 31.379 4.05 0.198
RPM 2 202.3 202.32 101.161 13.07 0.071
2
Residual 2 15.48 15.48 7.742
Error
 Total 8 338.8
4
Table 6.5 Analysis of Variance for SN ratios

ANOVA (Analysis of Variance) is a statistical method used to compare means between groups
in table 6.5. The "Source" column lists the factors being studied, such as temperature, time, and
residual error. "DF" (Degrees of Freedom) indicates the number of independent pieces of
information in the data set. "Seq SS" (Sequential Sums of Squares) measures the variability
between groups. "Adj SS" (Adjusted Sums of Squares) is an adjusted version of Seq SS that
considers the number of groups. "Adj MS" (Adjusted Mean Squares) is calculated by dividing
Adj SS by its corresponding DF. The "F" (F-statistic) compares the variances between groups
with the variance within groups. The "P" (p-value) indicates the probability of observing a test
statistic as extreme as the one calculated, assuming the null hypothesis is true, with a lower p-
value suggesting a more significant effect. From the table, temperature and time have p-values
greater than 0.05, indicating no significant effect on SN ratios, while the p-value for RPM is less
than 0.05, suggesting a significant effect on SN ratios. However, this table provides only partial
ANOVA analysis. A complete analysis would include information on the total sum of squares
and total degrees of freedom.

S.No. Initial Length Final Length Percentage Change

1 40.34 40 0.842836
2 40.63 40.5 0.319961
3 40.6 40.2 0.985222
4 40.02 39.56 1.149425
5 40.4 40.5 -0.24752
6 40.2 40 0.497512
7 40 39.8 0.5
8 40.2 40 0.497512
9 39.9 39.86 0.100251

Table 6.6 Measurements of length before and after smoothing

The table 6.6 shows the measurements of the change in length of nine objects. "S.No." likely
represents a serial number for each object. "Initial Length" denotes the original length of each
object, while "Final Length" indicates the length of each object after a measurement or process.
"Percentage Change" reflects the difference between initial and final lengths, expressed as a
percentage, with positive values indicating an increase and negative values indicating a decrease.
Additional observations reveal that the percentage changes are small, indicating no significant
length changes. Moreover, more objects exhibit a slight decrease in length compared to those
that show an increase.

Figure 6.4 Main effects plot for SN ratios

The figure 6.4 presents data from a signal-to-noise (SN) ratio experiment, where the SN ratio
compares the strength of a desired signal to background noise, with a higher ratio indicating a
stronger signal and lower noise level. The graph specifies that lower SN ratios are preferable in
this experiment, as indicated by the caption "Smaller is better" next to "Signal-to-noise." The
graph illustrates the effects of temperature, time, and RPM (revolutions per minute) on the SN
ratio. It shows that as temperature increases, the SN ratio generally decreases across all three
RPM levels. However, the impact of time on the SN ratio appears more complex and varies
depending on the temperature.

Seq
Source DF SS Adj SS Adj MS F P
Temperature 2 20.362 20.362 10.181 9.19 0.098
 Time 2 11.956 11.956 5.978 5.40 0.156

RPM 2 7.725 7.725 3.862 3.49 0.223

Residual 2 2.215 2.215 1.107

Error
Total 8 42.258

Table 6.7: Analysis of Variance for SN ratios

The table 6.7 titled "Analysis of Variance for SN ratio" displays the relationship between
temperature, the time, RPM and Residual Error it takes for a signal to reach a certain noise SN
ratio.

6.1.3 Width Measurement and Analysis

S.No. Initial Width Final Width Percentage Change
1 30.22 29.9 1.058901
2 30.6 30.4 0.653595
3 30.6 30 1.960784
4 29.9 29.62 0.936455
5 30.5 30.54 -0.13115
6 30.22 30.2 0.066181
7 29.84 29.78 0.201072
8 30.2 30.42 -0.72848
9 29.82 29.82 0

Table 6.8 Measurements of width before and after smoothing

The table 6.8 appears to be a dataset containing measurements of width, comprising four
columns: "S.No.," "Initial Width," "Final Width," and "Percentage Change." "S.No." likely refers
to a serial number, a unique identifier for each data point. "Initial Width" contains the initial
width measurement, possibly the first recorded value, while "Final Width" lists the final width
measurement, possibly the last recorded value. "Percentage Change" shows the percentage
change between the initial and final widths, with positive values indicating an increase, negative
values indicating a decrease, and a value of 0 indicating no change. Overall, the table seems to
summarize changes in width over time or between different conditions.
 Figure 6.5: Main Effects Plot for SN ratios
The figure 6.5 illustrates how the signal-to-noise ratio (SN) varies with different factors, namely
temperature, time, and RPM (revolutions per minute). The SN ratio is a measure used to
compare the strength of a desired signal to the level of background noise, with lower SN ratios
being better as they indicate less background noise relative to the signal. The graph likely shows
how the SN ratio changes at different temperature levels, indicating that temperature is a factor
influencing the SN ratio. Time is another factor that may affect the SN ratio, with the graph
potentially showing how different durations impact it. Additionally, RPM, representing
revolutions per minute, could relate to speed or rotation, with the graph indicating how different
RPM settings affect the SN ratio. The data representation in the graph appears to show the
average SN ratio for each factor level, such as at 50 degrees Celsius, where the average SN ratio
is indicated by a circle on the blue line. The vertical axis likely represents the range of SN ratios
from 1 to 9, with lower values indicating better signal quality. It is noted that the actual SN ratio
values are not displayed on the graph, only the averages for each factor level.

Adj Adj
Source DF Seq SS SS MS F P
Temperature 2 13.249 13.24 6.624 6.37 0.136
 9
Time 2 18.790 18.79 9.395 9.04 0.100
0
RPM 2 113.277 113.2 56.638 54.5 0.018
77 0
Residual 2 2.078 2.078 1.039
Error
Total 8 147.394
Table 6.9: Analysis of Variance for SN ratios

The table 6.9 represents a statistical analysis known as ANOVA, which compares different
groups to determine if their differences are likely due to chance. The "Source" column indicates
what is being compared, such as temperature, time, and RPM. "DF" (Degrees of Freedom) is a
number related to the amount of data being analyzed. "Sum of Squares (SS)" is a measure of
how spread out the data is, while "Mean Squares (MS)" is the average of the sum of squares. The
"F-value" is a number used to compare groups, and the "P-value" indicates if the differences are
likely due to chance. The ANOVA results suggest that RPM has the most significant impact on
the signal-to-noise ratio (SN ratio), followed by temperature and time. However, the effect of
temperature appears less straightforward and may require further investigation.

6.2 SHRINKAGE PHENOMENON

ABS thermoplastic shrinks when it cools after being extruded, which is compensated for by
FDM software. However, even with this compensation, small deviations can occur in complex
designs. This is because the layers of ABS material create internal stresses during cooling,
leading to deformation and additional shrinkage.

Three critical dimensions were chosen for study because of their complex shapes and sloping
profiles. The head diameter, for example, varies from base to tip, leading to errors in extrusion
and a smaller actual diameter compared to the CAD design. The neck profile has intricate
features with varying thickness and curved edges, causing errors in extrusion and resulting in a
thicker stem. Seams were also observed on the neck section of parts fabricated horizontally,
which affected dimensional accuracy.

The orientation angle during fabrication also affects the dimensional accuracy. Parts fabricated
horizontally (0° orientation) had non-uniform surface roughness and stair-stepping, while parts
fabricated vertically (90° orientation) had more uniform surfaces. This difference in orientation
affects how the material is deposited and the surface roughness, which in turn affects the
dimensional accuracy of FDM parts.

6.2.1 SHRINKAGE DUE TO LAYER RE-SETTLEMENT

After vapor smoothing, the ABS plastic flows and settles in the low areas, reducing surface
roughness. This happens because molten plastic naturally covers less surface area, which is
easier on flat surfaces. Vapor smoothing also temporarily melts the upper layers of the ABS
plastic due to the lower glass transition temperature caused by the hot chemical vapors. This
melting causes the peaks of the material to be lowered, leading to shrinkage in FDM replicas and
improving the accuracy of linear dimensions. Vapor smoothing improves the dimensional
accuracy of neck and stem sections by reducing shrinkage. However, it negatively affects the
dimensional accuracy of the head diameter, which is already produced slightly smaller by FDM.
This is because the smoothing process causes further shrinkage.

The smoothing process can result in both positive and negative effects on dimensional accuracy,
depending on the part feature. The process involves the downward movement (shrinkage) of
upper plastic layers at a micro-level, which reduces the overall dimensions of the replicas.
To avoid overheating, replicas must be cooled immediately after smoothing. The importance of
post-cooling is highlighted to prevent any potential issues. Overall, vapor smoothing reduces
surface roughness but can also cause deviations in dimensions, particularly in radial dimensions
like the head diameter.
 Figure 6.6 Deviations in (a) linear dimension before VS (b) linear dimension after VS
(c) radial dimension before VS (d) radial dimension after VS

After vapor smoothing, the settling of layers and smoother surfaces improve the
accuracy of Coordinate Measuring Machines (CMM). This improvement is shown by lower
standard deviation and standard error values after smoothing. The CMM probe can
accurately measure data points on the smooth surface with minimal error.

(a) (b)

Figure 6.7: SEM micrograph (a) before VS (b) after VS

SEM micrograph shows the layer shrinkage phenomenon with actual images. Before vapor
smoothing, you can clearly see the side view of a semi-circular profile with surface roughness.
After vapor smoothing, a smooth layer is observed because the semi-molten plastic material fills
in the empty spaces in valleys, reducing part dimensions and improving finishing.

6.2.2 REDUCTION IN STAIR-STEPPING

The stair-stepping phenomenon can be explained by the fundamental

manufacturing principle of FDM as discussed in chapter 2. During fabrication of slant
sections, the extrusion nozzle has to be raised periodically which generates a stair like
profile. The stair-stepping on the stem section of the composite replica can be
observed in Figure 6.8. The stair-steping deteriorates both the surface finish and
dimensional accuracy of parts.

Stairstepping

Figure 6.8 Stair-stepping visible on stem section of replica observed after fabrication

The comparison between original CAD dimension and actual surface with
stair-stepping has been visualized in Figure 6.8. After fabrication under FDM (before
vapor smoothing) the significant difference between CAD and actual surface is
visible. As the smoothing time is increased, the material reflowing rate increases
which decreases the height of the stair-step. Similar results have been acquired for
Taguchi
and ANOVA analysis of stem thickness. The percentage deviation decreases with increase
in smoothing time.

Figure 6.9 Stair-stepping reduction phenomenon

The large stair-step is visible in Figure 6.9a which depicts the condition of
surface before smoothing. After smoothing for 10s, the material transport occurs due
to partial melt down which reduces the height of stair step (Figure 6.9b). The replicas
are immediately hanged (post-cooled) in cooling chamber after smoothing which
results in freezing of plastic material. This reduces the height of stair step and also the
part dimensions which explains the reason for shrinkage in replicas after vapour
smoothing. The condition of stair step after smoothing for 15 and 20 seconds has been
shown in Figure 6.9c and 6.9d respectively. The increase in smoothing time further
results in greater reduction in amplitude of stair step. After vapour smoothing for
twenty seconds, the minute stair-step is viewed which validated the phenomenon
explained earlier. The SEM micrographs of stair-step have been acquired to validate
the smoothing phenomenon and dimensional changes occurring during finishing
 process (Figure 6.9). The difference between the magnitude of the stair-step before and
after smoothing can be clearly differentiated.

Large step
Large step

(a) (b)
Figure 6.10 SEM micrograph of stair-step (a) before VS (b) after VS

Figure 6.11 Normal Probability Plot for Length

The figure 6.11 you shared is a graph used to assess if data follows a normal distribution, a
common pattern in many natural phenomena. The x-axis displays the differences between
observed and predicted values, while the y-axis shows the percentage of data points below a
certain difference. A straight line on the graph indicates that the data is normally distributed.
However, slight deviations from this line, especially in the tails of the graph, suggest that the
data may not be perfectly normal. In this graph, there are slight curves in the tails, indicating
some deviations from normality. Overall, the data doesn't appear to be extremely non-normal,
but it's also not perfectly normal.

Figure 6.12 Normal Probability Plot for Width

In the figure 6.12, the x-axis represents the differences between observed and predicted values,
while the y-axis indicates the percentage of data points below a certain difference. A straight line
on the graph would suggest that the data follows a normal distribution. However, the presence of
slight curves in the tails indicates some deviations from normality. While these deviations are
not extreme, they suggest that the data is not perfectly normal. Overall, the data appears to be
reasonably close to a normal distribution, but it is not entirely normal.
 Figure 6.13 Normal Probability Plot for Thickness

In figure 6.13 , the x-axis shows the differences between observed and predicted thicknesses,
while the y-axis shows the percentage of data points below a certain difference. A straight line
on the graph suggests the data is normally distributed. However, slight deviations from this line,
especially in the tails of the graph, indicate the data may not be perfectly normal. In graph, there
are slight curves in the tails, indicating some deviations from normality. Overall, the data doesn't
seem extremely non-normal, but it's not perfectly normal either.
 Figure 6.14 Normal Probability Plot for Diameter

In figure 6.14, the x-axis shows the differences between observed and predicted diameters, while
the y-axis shows the percentage of data points below a certain difference. A straight line on the
graph suggests the data is normally distributed. However, slight deviations from this line,
especially in the tails of the graph, indicate the data may not be perfectly normal. In graph, there
are slight curves in the tails, indicating some deviations from normality. Overall, the data doesn't
seem extremely non-normal, but it's not perfectly normal either.
6.3 MULTI-RESPONSE OPTIMIZATION AND CONFIRMATORY
EXPERIMENT

Parameter Surface roughness Surface hardness Length

Temperature 70 60 50
Time(min) 20 40 40
Air flow(RPM) 2670 2670 2490
Parameter Surface roughness Surface hardness Length
Temperature 70 60 50
Time(min) 20 40 40

Table 6.10: Optimum parameter settings for each part feature

The table 6.10 presents measurements from an experiment, divided into two sections, each
starting with the "Parameter" header. The first column lists the properties measured, including
Surface roughness, Surface hardness, Length, Temperature, Time (min), and Air flow (RPM).
The subsequent columns contain the recorded values for each property. Each group of
measurements begins with the "Parameter" header, and there are two sets of measurements in the
table. For each property, the first row names the property, and the following rows display the
measured values. Overall, the table depicts the results of an experiment measuring various
properties under different conditions, as indicated by the two sets of measurements.
Parameter Goal Lower Limit Upper limit Weight Imporatance
Temperature Constrain to 40 70 1 1
region
Time(min) Constrain to 20 60 1 1
region
Air flow(RPM) Constrain to 1880 2670 1 1
region
SN Ratios Maximize 36.5150 39.8740 1 1
%∆Ra
SN Ratios %∆Hd Maximize 16.6502 36.8472 1 1
SN Ratios %∆L Maximize 1.39121 4.08240 1 1
Length
SN Ratios %∆W Maximize 0.1137 12.6902 1 1
Width
SN Ratios %∆T Maximize -23.0745 -4.0061 1 1
Thickness
SN Ratios %∆D Maximize -9.4843 8.8739 1 1
Diameter

Table 6.11: Constraints and goals set for multi-response optimization

The table 6.11 outlines the parameters and goals for an experiment focused on optimizing
multiple signal-to-noise (SN) ratios. The "Parameter" column lists the factors controlled or
measured in the experiment, including Temperature, Time (min), Air flow (RPM), and various
SN ratios. The "Goal" column specifies desired outcomes for each parameter, such as "Constrain
to region" to keep within a defined range (with lower and upper limits) or "Maximize" to achieve
the highest value. For parameters constrained to a region, the table includes "Lower Limit" and
"Upper Limit" columns indicating the minimum and maximum acceptable values, respectively.
The "Weight" column indicates the relative importance of each parameter in the experiment,
with all weights set to 1, suggesting equal importance. Similarly, the "Importance" column
provides another rating of importance, also set to all 1, indicating equal importance for all
parameters. Overall, the table sets up an experiment where researchers aim to control
environmental factors (temperature, time, airflow) within specific ranges and maximize multiple
SN ratios. By controlling these factors, researchers aim to find the best combination that
maximizes SN ratios, with the equal weight and importance values suggesting that all parameters
are equally crucial for the experiment's success.
.

Parameter Value
Temperature 45
Time(min) 32
Air flow(RPM) 2654
SN Ratios 39.5767
%∆Ra
SN Ratios %∆Hd 22.0934
SN Ratios %∆L Length 2.46321
SN Ratios %∆W Width 1.4562
SN Ratios %∆T Thickness -2.1323
SN Ratios %∆D Diameter 3.8739
Desirability 0.564840

Table 6.12: Optimum settings of parameters and predicted SN ratio

The table 6.12 displays the results from a single trial of an experiment,
showing the measured values of different parameters. The "Parameter"
column lists the properties measured, such as Temperature, Time (min), Air
flow (RPM), SN Ratios (%∆Ra, %∆Hd, %∆L Length, %∆W Width, %∆T
Thickness, %∆D Diameter), and Desirability. The "Value" column shows the
numerical measurements for each parameter during this trial, while the
"Desirability" column indicates a score (0.564840) summarizing how well the
trial met the experiment's overall goals. Overall, the table presents the
results of one experimental run, showing the controlled factors
(temperature, time, airflow) and measured properties, with the desirability
score summarizing the trial's success in meeting the experiment's goals.
6.3 MATHEMATICAL MODELING OF RESPONSE

Head Diameter Neck thickness Stem thickness

% IT % IT % IT
Ex Readin Readin Reading
ΔDa Grad ΔDa Grad ΔDa Grad
p. gs gs s
e e e
1 12.0102 0.35 IT9 7.7621 0.21 IT8 6.6498 0.44 IT9
2 12.0099 0.35 IT9 7.7626 0.21 IT8 6.6474 0.41 IT9
3 12.0089 0.36 IT9 7.7622 0.21 IT8 6.6488 0.43 IT9
4 12.0095 0.35 IT9 7.7610 0.19 IT8 6.6492 0.43 IT9
5 12.0124 0.33 IT9 7.7593 0.17 IT8 6.6482 0.42 IT9
6 12.0115 0.34 IT9 7.7595 0.17 IT8 6.6475 0.41 IT9
7 12.0088 0.36 IT9 7.7615 0.20 IT8 6.6487 0.43 IT9
8 12.0094 0.35 IT9 7.7598 0.18 IT8 6.6472 0.40 IT9

Table 6.13 Percentage deviations of part features at optimum parameter settings

The table 6.13 presents the results of eight measurements taken on different parts of an object.
The first three columns display the measured values for each part, including Head Diameter,
Neck Thickness, and Stem Thickness. The following three columns show the grade assigned to
each measurement based on unspecified criteria, labeled as Neck Grade, Stem Thickness %, and
Stem Grade. The last three columns, namely Stem Thickness, Exp. Diameter, and Exp.
Thickness %, likely relate to expected values for these parts, although the precise meaning of
these columns is unclear without additional context. Each row in the table represents a specific
measurement, with corresponding values for the measured parameters and their assigned grades.
6.3.1 SELECTION OF PARAMETERS FOR MATHEMATICAL
MODEL
The Buckingham Pi theorem was used for mathematical modeling in this
experiment. Since the output parameter is a percentage deviation, it can't
be included because it lacks valid units. Instead, the deviation after vapor
smoothing (∆𝐷𝑎) is used as the response, with fundamental units of M0 L1 T0. This
deviation is calculated as:

∆𝐷𝑎 = Original (CAD) dimension – Actual dimension

The orientation angle is important but can't be an input parameter because

it lacks valid units. However, it significantly impacts the percentage
deviation before vapor smoothing (∆𝐷𝑏), which can be expressed as:

∆𝐷𝑏 ∝ Orientation angle

The initial deviation is directly related to the final deviation, so:

∆𝐷𝑎 ∝ ∆𝐷𝑏

Therefore, the percentage deviation before smoothing is used as an input parameter.

The input parameters, their units, and dimensions assumed to create the dimensional model are:

- Initial deviation (∆𝐷𝑏) in mm: M0 L1 T0

- Density (ρ) in g/mm³: M1 L-3 T0

- Pre-cooling Time (TPC) in minutes: M0 L0 T1
- Smoothing Time (TS) in seconds: M0 L0 T1
- Post-cooling Time (TPT) in minutes: M0 L0 T1
6.4 FORMULATION OF MATHEMATICAL MODEL

The percentage deviation depends on all the input parameters:

∆𝐷𝑎 = f(∆𝐷𝑏, ρ, TPC, TS, TPT) (6.6)
Here, n is 6 and m is 3, resulting in three (n - m = 3) independent dimensionless Π terms: Π 1, Π2,
and Π3. The initial deviation, density, and pre-cooling time are the repeating variables included
in each Π term, while the other three parameters appear once in each term. According to the
Buckingham Pi Theorem, each parameter represents at least one dimension (mass, length, or
time).
Π1 = ∆𝐷𝑎 (∆𝐷𝑏)a1 (ρ)b1 (TPC)c1 (6.7)
Π2 = TS (∆𝐷𝑏)a2 (ρ)b2 (TPC)c2 (6.8)
Π3 = TPT (∆𝐷𝑏)a3 (ρ)b3 (TPC)c3 (6.9)
Since Π terms are dimensionless, the dimensions are substituted into the equations and equated
to zero to find the exponents ai, bi, and ci.
For Π1:
Π1 = (M0 L T0) (L)a1 (ML-3)b1 (T)c1
Equating terms in LHS equal to M0 L0 T0 gives b1 = 0, c1 = 0, -3b1+a1+1=0 and a1 = -1. Thus:
Π1 = ∆𝐷𝑎 (∆𝐷𝑏)-1 (6.10)
For Π2:
Π2 = M0 L0 T (L)a2 (ML-3)b2 (T)c2
Equating to M0 L0 T0 gives b2 = 0, 1+c2 = 0, c2=-1, a2-3b2=0 and a2 = 0. Thus:
Π2 = TS / TPC (6.11)
For Π3:
Π3 = T(L)a3 (ML-3)b3 (T)c3
Equating to M0 L0 T0 gives b3 = 0, c3 = -1, 1+c3=0 and a3 = 0. Thus:
Π3 = TPT / TPC (6.12)
The ultimate relationship is:
Π1 = f(Π2, Π3)
Inserting values from Equations 6.10, 6.11, and 6.12 in ultimate relationship, we get
∆ Da Ts . T PT
=f ( )
∆ Db T PC .T PC
∆ Da
=f ¿) (6.13)
∆ Db
Now, we know that
𝐷𝑎 = original (CAD) dimension – actual dimension after smoothing
And, 𝐷b = original (CAD) dimension –actual dimension before smoothing
So, by dividing and multiplying the equation (6.13) by:
(Original CAD dimension× 100)
We get,
∆ Da ×(original CAD dimension ×100) K N T S T PT
=
∆ Db ×(original CAD dimension ×100) T PC 2
Rearranging we get:
∆ Da ×100
/ ∆ D b × 100
(original CAD dimension) K T T
= N S PT
( original CAD dimension) T PC 2

% ∆ Da
=¿)
% ∆ Db
% ∆ Db K N T S T PT
% ∆ Da=( ) (6.14)
T PC 2
Where KN is a constant of proportionality

6.5 EXPERIMENTATION FOR EQUATION DEVELOPMENT

The table outlines the constraints and goals for an experiment, listing parameters that will be
controlled or measured. Here's a simpler breakdown:
Parameter: Lists the factors being controlled or measured, including:
- Temperature
- Time (min)
- Air flow (RPM)
- SN Ratios (%∆Ra, %∆Hd, %∆L Length, %∆W Width, %∆T Thickness, %∆D
Diameter)
Goal: Specifies the desired outcome for each parameter:
- "Constrain to region": Keep within a specific range (lower and upper limits).
- "Maximize": Aim to achieve the highest possible value.
Lower Limit: Minimum acceptable values for parameters that need to be constrained.
Upper Limit: Maximum acceptable values for parameters that need to be constrained.
Weight: Indicates the relative importance of each parameter in achieving the experiment's goal.
Importance: Another rating of importance, with all values being 1, suggesting equal importance
for all parameters.

Overall, the table sets the conditions for an experiment aimed at

optimizing multiple SN ratios (%∆Ra and %∆Hd) by controlling factors
like temperature, time, and airflow within specific ranges, and
maximizing the SN ratios.

1.8
1.6
1.4
Per
1.2cen
tag
1e
0.8dev
iati
0.6on
0.4
0 5 10 15 20 25 30 35
Smoothing time

Figure 6.15 Percentage deviations at variable smoothing time

KN = 0.054 (TS)2 – 0.243 (TS) + 5.364 for head diameter

KN = -0.003 (TS)2 – 0.081 (TS) + 6.238 for neck thickness

KN = 0.003 (TS)2 – 0.269 (TS) + 5.341 for stem thickness

Thus, to predict deviation in desired part feature, the value of KN for part feature can be inserted
in generalized equation 6.14. The final equation for prediction of percentage deviation in head
diameter can be written as:

% ∆ Db (0.054 T S 2 – 0.243T S +5.364 )T S T PT

% ∆ Da=( ) (6.15)
T PC 2
 6.5.1 VALIDATION OF MATHEMATICAL MODEL

Figure 6.16 I Chart and Moving Range Chart

The provided figure 6.16 appear to be part of a statistical process control (SPC) application,
commonly used to monitor processes over time and identify significant variations. The Moving
Range Chart displays the differences between consecutive measurements (MR) on the y-axis,
fluctuating around 2. It has an upper control limit (UCL) of 4.31 and a lower control limit (LCL)
of 0. The Individual Value Chart, on the other hand, shows the values of a single parameter
across the samples, labeled as "Value" on the y-axis. The specific parameter being measured is
not labeled. The values range from about -1 to 5, with a UCL of 5.38 and an LCL of -1.63.
Control limits such as the UCL and LCL are set based on historical data and define the expected
range of variation. If a data point falls outside these limits, it may indicate that the process is out
of control and requires further investigation.

Figure 6.17 Diameter data analysis (a) X-chart (B) R-chart (c) Normal probability plot
The figure 6.17 you described is a normal probability plot, which is used to check if a dataset
follows a normal distribution (bell curve). The horizontal axis is labeled "Normal Order
Statistics," showing theoretical quintiles of a normal distribution, while the vertical axis is
labeled "Data," displaying your dataset values from lowest to highest. The red line represents the
theoretical normal distribution. Interpretation of the plot suggests that if the data points align
closely with a straight line, the data is likely normally distributed. However, in your plot, the
data points deviate slightly from the red line, especially at the ends, indicating that your data
might not be perfectly normal. It’s important to note that for some statistical tests, minor
deviations from normality are acceptable. It's advisable to consult with a statistician to determine
if this non-normality affects your analysis. Additionally, alternative plots such as histograms and
Q-Q plots can also be used to assess normality. In summary, while the normal probability plot
suggests your data might not be perfectly normal, the impact of this non-normality depends on
the specific context and purpose of your analysis.

Figure 6.18 I Chart and Moving Range Chart

The figure 6.18 you described includes an Individual (I) Chart and a Moving Range (MR) Chart,
commonly used in statistical process control (SPC) to monitor processes over time. The
Individual Chart plots individual values of a parameter over time, with the y-axis labeled
"Value" and the x-axis labeled "Sample," numbered from 1 to 9. The values fluctuate around 50,
suggesting this might be the average or target value for the process. The Moving Range Chart,
on the other hand, plots the differences between consecutive measurements. The y-axis is labeled
"Moving Range (MR)," and the x-axis is labeled "Sample." The moving range values fluctuate
around 5, with an upper control limit (UCL) of 14.28 and a lower control limit (LCL) of 0.
Control Limits (UCL and LCL) are set based on historical data and define the expected range of
variation. Points outside these limits suggest that the process might be out of control and require
investigation. In the image, all data points are within the control limits, indicating that the
process seems stable. Overall, the I-MR chart suggests that the process is currently stable and
under control. However, it's important to keep monitoring the chart to detect any future changes
that could indicate instability.

Figure 6.19 Thickness data analysis (a) X-chart (B) R-chart (c) Normal
probability plot

The figure 6.19 illustrates how changing the levels of Factor A and Factor B influences the
response variable. Although the specific levels are not shown, the contour lines connect points
with the same value of the response variable. Darker colors indicate higher response values,
while lighter colors indicate lower values. Interpretation of the plot suggests that the response
variable is highest in the upper left and lower right corners, and lowest in the center. This pattern
indicates a potential interaction effect between Factor A and Factor B, meaning the effect of one
factor on the response depends on the level of the other factor. Response Surface Methodology
(RSM), represented by this plot, is used in various fields to optimize processes. For instance, it
can help identify the best combination of factors for a chemical reaction or product quality.
Overall, this plot provides a visual representation of how two factors influence a response
variable, offering insights for process optimization.
 Figure 6.20 I Chart and Moving Range Chart

The provided figure 6.20 information describes an Individual Chart (I-Chart) and a Moving
Range Chart (MR Chart) used in statistical process control (SPC) to monitor processes over
time. The I-Chart displays individual values of a parameter over time, with the y-axis labeled "I"
and the x-axis labeled "Sample" from 1 to 11. The values fluctuate around 3, suggesting this
might be the average or target value for the process. The MR Chart shows the difference
between consecutive measurements, with the y-axis labeled "MR" and the x-axis also labeled
"Sample." The moving range values fluctuate around 0.5, with control limits at 1.08 (UCL) and
0 (LCL).Interpretation of the charts indicates that all data points are within the control limits,
indicating a stable and predictable process. The average or target value appears to be around 3
based on the I-Chart, and the variability between consecutive measurements is relatively small,
as seen in the MR Chart. Overall, the combined I-MR chart suggests that the process is currently
stable and under control. Continued monitoring of the chart over time is important to detect any
future changes that may indicate process instability.
 Figure 6.21 Length data analysis (a) X-chart (B) R-chart (c) Normal
probability plot

The figure 6.21 you're describing is a normal probability plot, which is commonly used to assess
if a dataset follows a normal distribution. In this plot, the horizontal axis represents theoretical
quantiles of a normal distribution, while the vertical axis represents the residuals, which are the
differences between observed and expected values. Interpreting the plot, if the points align
closely with a straight line, it suggests that the residuals are normally distributed. However, if the
points deviate from the line, especially for extreme values, it indicates potential non-normality in
the data. It's important to note that non-normality may not be critical depending on the analysis
being conducted. In cases where non-normality is a concern, other methods such as histograms
and Q-Q plots can also be used to assess normality.

Figure 6.22 I Chart and Moving Range Chart

The figure 6.22 you described is a normal probability plot, commonly used to assess if a dataset
follows a normal distribution. The horizontal axis represents theoretical quantiles of a normal
distribution, while the vertical axis represents the residuals, which are the differences between
observed and expected values. Interpreting the plot, if the points align closely with a straight
line, it suggests that the residuals are normally distributed. However, if the points deviate from
the line, especially for extreme values, it indicates potential non-normality in the data. Non-
normality may not be critical depending on the analysis being conducted. In cases where non-
normality is a concern, other methods such as histograms and Q-Q plots can also be used to
assess normality.

Figure 6.23 Width data analysis (a) X-chart (B) R-chart (c) Normal
probability plot

The figure 6.23 you're referring to is a Pareto chart, a combination of a bar graph and a line
graph. Here's a breakdown of its components and interpretation: Bar Graph: The bars represent
the frequency or count of each category, with the tallest bar corresponding to the most frequent
category (Defect 2 in this case). This part of the chart helps visualize the relative importance of
each category in terms of frequency of occurrence. Line Graph: The line graph shows the
cumulative percentage of the frequency for each category. It indicates the total contribution of
each category to the overall defects. The steeper the slope of the line, the more impactful the
category is in contributing to the total defects. Pareto Principle: The Pareto principle, or 80/20
rule, is often applied to Pareto charts. It suggests that roughly 80% of effects come from 20% of
causes. In the context of quality control, this means that a small number of categories (the "vital
few") contribute to the majority of defects, while the majority of categories (the "trivial many")
contribute to a smaller portion of defects. Interpretation: In your chart, Defect 2 is the most
common, followed by Defect 1 and Defect 3. These categories likely account for a large portion
of the total defects. By focusing efforts on addressing these critical defect categories, quality
control processes can be significantly improved. Overall, the Pareto chart helps prioritize efforts
by highlighting the most critical defect categories, allowing organizations to allocate resources
more effectively for maximum impact on quality improvement.