16 - CHAPTER 3 Thesis - 3
16 - CHAPTER 3 Thesis - 3
16 - CHAPTER 3 Thesis - 3
CHAPTER 3
METHODOLOGY
3.1 INTRODUCTION
This chapter was discussed about overall work flow progress of the project
including the flow of experiment done using appropriate apparatus, the information
about the material had been used, designing of the experiment, the experiment set up
which is involving machining process, and also the method that had been used to
interpret the data collected.
Apart from that, the appropriate apparatus and the experiment set up was well
prepared in order to have a smooth handling during grinding process so that the
objective of the project is achieve successfully. Therefore, the data collection will be
analyze clearly and interpret it using a suitable way to evaluate and compare it with the
previous research.
27
The most typical materials for the grinding wheel that have been used in
mechanical laboratory of University Malaysia Pahang are Al2O3 (figure 3.3) and SiC
(figure 3.4). The material of Al2O3 is not so hard and fragile as SiC. It is suitable for
grinding the metals with high anti-tensile strength such as softer metal, forge iron and
bronze. For the SiC, the crystals are very fragile. Regarding to this kind of property, it is
suitable for grinding the metals with high anti-tensile strength too such as steel, marble
and glass.
When selecting a grinding wheel, there are several factors that had been
considered to conduct the experiment where we use the ISO standard for grinding
selection. The wheel specifications are:
dressing process was performed to the wheel in each stage of grinding process by using
diamond wheel dresser in figure 3.5.
Most grinding machines are equipped with coolant systems. In my final year
project 1, the water based coolant had been used which is figure 3.6. The water based
coolant mixture contains 60% of water and 40% ethylene glycol. The coolant was
installed to the machine as in figure 3.7 and is directed over the point of contact
between the grinding wheel and the work. This prevents distortion of the workpiece due
to uneven temperatures caused by the cutting action. In addition, coolant keeps the chips
washed away from the grinding wheel and point of contact, thus permitting free cutting.
Clear water may be used as a coolant, but various compounds containing alkali are
usually added to improve its lubricating quality and prevent rusting of the machine and
workpiece. For grinding aluminum and its alloys, a clear water coolant will produce
fairly good results.
32
3.2.5 Workpiece
The material that had been used as workpiece in this project is Haynes 242 alloy.
HAYNES 242 alloy is an age-hardenable nickel-molybdenum-chromium superalloy
that derives its good mechanical properties from a fine dispersion of metastable
Ni2(Mo,Cr) ordered body centered orthorhombic (Pt2Mo-type) domains/precipitates in
a face centered cubic γ matrix. The property of the alloy is high strength and resistance
to corrosion which are critical considerations in the selection of materials for plastics
production and processing equipment. It strength is twice the strength of traditional
HASTELLOY alloy up to 100°F (650°C). Figure 3.10 show the workpiece is prepared
in the form of 20mm thickness, 175mm length and 107mm width. It has yield strength
about 200ksi (1380Mpa), 9.06g/cm³ of density, reach the electrical resistance up to
122.0µohm. In addition, the dynamic modulus elastic of the material is 229 Gpa and
have a good thermal conductivity.
The experiments set up was begin with the surface grinding machine. The
machine was ensured that there is no systematic error and was in good condition
whether it is to be control manually or automatically. The crucial part that was need to
be observe is the emergency button was well functioned during the grinding process.
The control panel which is control the accuracy of feed rate at the x, y and z-axis was
ensured so that it will give the good accuracy of the experiments. The material that was
used in this project is Haynes 242 which 175mm in length and 107mm in width with
thickness of 2mm. The material is a non-magnetic to the table of the machine resulting
it moves randomly on the table speed. The magnetic vice was used to clamp the
workpiece to make sure it was stick and does not even an inch during the grinding
process. Since the workpiece had a quite long dimensions in the length, two vice was
use to support the workpiece on the flat surface on the table speed. The vice was locked
tightly until the workpiece cannot move at all.
The workpiece now ready to under goes into the grinding process using the SiC
wheel with the constant table speed which is 200rpm. The table speed was measured by
using the tachometer in the laboratory. First stage of the experiment is wanted to flat the
original surface so that the good results obtained. The new flat surface was obtained and
the value of Ra before the actual grinding process was measured using Mahr
Perthometer S2 equipment. It is used to measure the surface roughness of the material.
All the data was collected and placed in the table properly. The reading was taken about
three times to make sure the accuracy of the data that was collected.
When the value of Ra before was taken, the real experiment was run by setting
the value of the depth of cut which is 5µm. However, the dressing operation of the
wheel was done before the real experiments and the wheel starts to grind the surface of
the workpiece. The operation was done by using a diamond dresser. After the dressing
operation was done, the diameter of the wheel before its grind was taken by using the
vernier caliper in tha laboratory. A single pass process was taken place where the
Haynes 242 was grinds only one passes through surface with the addition of the water
based coolant during the processes. Once it is finished grinds, the workpiece was
37
bringing to the metrology laboratory to measure it new surface roughness using Mahr
Perthometer S2. This is the new value of Ra where it was taken after the grinding
processes were done. The diameter of the wheel after the grinding process also taken
and was tabled.
Flatting the new surface of the material had been done after for every single-
pass grinding process. The value of Ra of the material after the flatting process will be
taken again. A value of Ra now, becomes the value or Ra before the grinding process
with 7 µm of the depth of cut. The experiment were continued and was repeated using
different value depth of cut which are 7, 9, 11, 13, 15, 17, 19, 21 µm. The experiments
were conducted repeatedly using another wheel which is Al2O3 wheel. After two set of
experiment had been done, the experiment is continued by using multi-pass grinding
process with the same procedure as single-pass grinding process. All the data was
recorded and calculated where it is ready to be analyzes by the neural network analysis.
Networks with biases can represent relationships between inputs and outputs
more easily than networks without biases. A transfer function, consisting generally of
algebraic equations, may be linear or non-linear. The most basic and commonly used
ANN is the multi-layer perception (MLP). This consists of at least three or more layers,
an input layer, an output layer, and a number of hidden layers.
The experimental data set consist of nine values. The ANN model was trained
using 7 randomly selected data, accounting for 70% of the total data while the
remaining two data accounting for 15% for each, was utilized for testing of the network
performance. An Architecture design as figure 3.11 was done which are including three
elements, an input layer, a hidden layer, and an output layer network. There is one input
in the input layer, namely depth of cut. The output is the surface roughness and wheel
wear produced. But since the analysis was using the Levenberg-Marquadt training
algorithm, which the summation of the square error testing, it cannot analyzed the wheel
wear produced due to the smallest value obtained which it is not significance. Thus, the
analysis will concentrate only on the surface roughness as the output analysis.
39
Correlation: 0.999133
R-Squared: 0.999093
In this study, the ANN model is developed for predicting the value of surface
roughness with less error occur which indicate with the correlation coefficient and
coefficient of determination. To evaluate theaccuracy of the ANN predictions, each
graph is provided with a straight line indicating the perfect prediction. Each graph is
provided with a accuracy of the ANN predictions. Each graph are also is provided with
a positive line indicating the perfect predictions. The performance of the ANN
prediction was evaluated by a regression analysis between the predicted and the
experiment value as shown in figure 3.12 above. The comparisons were made using
only the values from the test data set, which was not introduced to the network during
the training process and was selected randomly from experimentally obtained data set.
Figure 3.11 shows a regression lines indicates if there is no linear relationship between
the variables or only a weak relationship, r will be close to 0. From the table 3.1 above,
the correlation coefficient is nearly to the 1. It is mean it has a strong positive
relationship between the variables.It is a statistical method used to measure two or more
variables are related and to determine whether a relationship between variables exists
and how strong is the linear relationship between the variables. The stronger the
relationship between variables, the more accurate the prediction is.